[svn-r876] Changes since 19981102
----------------------
./bin/snapshot
Made same fix as for the release script yesterday.
./src/H5D.c
./src/H5Dprivate.h
./src/H5G.c
./src/H5Gprivate.h
./src/H5Gpublic.h
./src/H5O.c
./src/H5Oprivate.h
./src/H5RA.c
./src/H5RAprivate.h
./src/H5T.c
./src/H5Tprivate.h
Improved object type checking. Instead of determining the
object type by trying to open each of the possible types, we
keep a table of associations between object type number (like
H5G_GROUP, H5G_DATASET, H5D_TYPE, and H5D_RAGGED) and an `isa'
function that returns true if the object header has the right
messages to make the object a particular type. This mechanism
also allows specialization of object types by permitting an
object to satisfy more than one `isa' function.
Added `isa' functions for groups, datasets, ragged arrays, and
committed data types.
./src/H5config.h.in
Added HAVE_STAT_ST_BLOCKS. I thought this had already been
added, but apparently not.
./tools/h5ls.c
Removed system include files since they're already included by
H5private.h and since I wasn't including them portably anyway.
By default, 1-byte integer types are printed as integer values
instead of ASCII characters. However, the `-s' or `--string'
command-line switch causes the data to be interpretted as
ASCII. String data types are always printed as character
data.
Ragged arrays are now identified as ragged arrays and h5ls
doesn't descend into the group automatically. This uses the
new object type specialization stuff.
./tools/h5tools.c
./tools/h5tools.h
Added the ability to print 1-byte integer types as either
ASCII or numeric data instead of always ASCII. The default is
to print as numeric data.
1998-11-06 04:28:34 +08:00
dnl Process this file with autoconf to produce configure.
1997-08-16 00:35:53 +08:00
dnl
2007-02-21 04:31:14 +08:00
dnl Copyright by The HDF Group.
2002-07-18 00:10:46 +08:00
dnl Copyright by the Board of Trustees of the University of Illinois.
dnl All rights reserved.
dnl
dnl This file is part of HDF5. The full HDF5 copyright notice, including
dnl terms governing use, modification, and redistribution, is contained in
dnl the files COPYING and Copyright.html. COPYING can be found at the root
dnl of the source code distribution tree; Copyright.html can be found at the
dnl root level of an installed copy of the electronic HDF5 document set and
dnl is linked from the top-level documents page. It can also be found at
2007-02-21 04:31:14 +08:00
dnl http://hdfgroup.org/HDF5/doc/Copyright.html. If you do not have
dnl access to either file, you may request a copy from help@hdfgroup.org.
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
dnl Initialize configure.
dnl
AC_REVISION($Id$)
2003-08-04 23:30:54 +08:00
AC_PREREQ([2.53])
2002-04-24 04:02:26 +08:00
dnl AC_INIT takes the name of the package, the version number, and an
dnl email address to report bugs. AC_CONFIG_SRCDIR takes a unique file
dnl as its argument.
dnl
dnl NOTE: Don't forget to change the version number here when we do a
dnl release!!!
dnl
2011-05-23 22:19:26 +08:00
AC_INIT([HDF5], [1.9.82], [help@hdfgroup.org])
2002-04-24 04:02:26 +08:00
AC_CONFIG_SRCDIR([src/H5.c])
2005-02-01 11:17:02 +08:00
AM_CONFIG_HEADER([src/H5config.h])
2002-04-24 04:02:26 +08:00
AC_CONFIG_AUX_DIR([bin])
2010-09-10 23:45:07 +08:00
AC_CONFIG_MACRO_DIR([m4])
2005-05-26 06:32:22 +08:00
dnl AM_INIT_AUTOMAKE takes a list of options that should be applied to
dnl every Makefile.am when automake is run.
AM_INIT_AUTOMAKE([foreign])
2011-04-15 03:45:57 +08:00
AM_SILENT_RULES([yes])
2002-04-24 04:02:26 +08:00
2005-02-03 04:59:46 +08:00
dnl AM_MAINTAINER_MODE turns off "rebuild rules" that contain dependencies
dnl for Makefiles, configure, src/H5config.h, etc. If AM_MAINTAINER_MODE
dnl is *not* included here, these files will be rebuilt if out of date.
dnl This is a problem because if users try to build on a machine with
dnl the wrong versions of autoconf and automake, these files will be
dnl rebuilt with the wrong versions and bad things can happen.
dnl Also, CVS doesn't preserve dependencies between timestamps, so
dnl Makefiles will often think rebuilding needs to occur when it doesn't.
dnl Developers should './configure --enable-maintainer-mode' to turn on
dnl rebuild rules.
AM_MAINTAINER_MODE
2008-05-15 13:12:00 +08:00
dnl Run post processing on files created by configure.
dnl src/H5pubconf.h:
dnl Generate src/H5pubconf.h from src/H5config.h by prepending H5_ to all
dnl macro names. This avoid name conflict between HDF5 macro names and those
dnl generated by another software package that uses the HDF5 library.
dnl src/libhdf5.settings:
dnl Remove all lines begun with "#" which are generated by CONDITIONAL's of
dnl configure.
1999-11-17 03:08:14 +08:00
AC_OUTPUT_COMMANDS([
2001-09-28 07:05:16 +08:00
echo "creating src/H5pubconf.h"
sed 's/#define /#define H5_/' <src/H5config.h |\
sed 's/#undef /#undef H5_/' >pubconf
if test ! -f src/H5pubconf.h; then
/bin/mv -f pubconf src/H5pubconf.h
elif (diff pubconf src/H5pubconf.h >/dev/null); then
2007-02-07 02:50:04 +08:00
rm -f pubconf
2001-09-28 07:05:16 +08:00
echo "src/H5pubconf.h is unchanged"
else
/bin/mv -f pubconf src/H5pubconf.h
fi
2008-05-15 13:12:00 +08:00
echo "Post process src/libhdf5.settings"
sed '/^#/d' < src/libhdf5.settings > libhdf5.settings.TMP
cp libhdf5.settings.TMP src/libhdf5.settings
rm -f libhdf5.settings.TMP
1999-11-17 03:08:14 +08:00
])
2002-04-24 04:02:26 +08:00
2005-06-21 01:33:22 +08:00
dnl It's possible to configure for a host other than the one on which
dnl configure is currently running by using the --host=foo flag.
dnl For machines on which HDF5 is often configured, it can be convenient
dnl to specify the name of the machine rather than its canonical type.
case $host_alias in
2005-12-08 04:45:42 +08:00
redstorm)
host_alias=x86_64-redstorm-linux-gnu
2005-06-21 01:33:22 +08:00
;;
esac
1997-08-16 00:35:53 +08:00
AC_CANONICAL_HOST
2002-07-18 00:10:46 +08:00
AC_SUBST([CPPFLAGS])
1997-08-16 00:35:53 +08:00
2005-09-17 05:26:17 +08:00
dnl H5_CFLAGS (and company) are for CFLAGS that should be used on HDF5, but
dnl not exported to h5cc (or h5fc, etc.)
AC_SUBST([H5_CFLAGS])
AC_SUBST([H5_CPPFLAGS])
2006-04-22 03:21:24 +08:00
AC_SUBST([H5_FCFLAGS])
2005-09-17 05:26:17 +08:00
AC_SUBST([H5_CXXFLAGS])
2011-04-15 05:21:59 +08:00
AC_SUBST([H5_LDFLAGS])
2005-09-17 05:26:17 +08:00
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
dnl AM_CFLAGS (and company) are for CFLAGS that should be used on HDF5,
dnl and WILL be exported to h5cc (or h5fc, etc) if set by configure.
AC_SUBST([AM_CFLAGS])
AC_SUBST([AM_FCFLAGS])
AC_SUBST([AM_CXXFLAGS])
AC_SUBST([AM_CPPFLAGS])
AC_SUBST([AM_LDFLAGS])
2006-04-27 05:19:21 +08:00
dnl Make sure flags are set to something (otherwise macros may set them later).
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CFLAGS="${AM_CFLAGS}"
AM_CXXFLAGS="${AM_CXXFLAGS}"
AM_FCFLAGS="${AM_FCFLAGS}"
AM_CPPFLAGS="${AM_CPPFLAGS}"
AM_LDFLAGS="${AM_LDFLAGS}"
2006-04-27 05:19:21 +08:00
CFLAGS="${CFLAGS}"
CXXFLAGS="${CXXFLAGS}"
FCFLAGS="${FCFLAGS}"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
CPPFLAGS="${CPPFLAGS}"
LDFLAGS="${LDFLAGS}"
2009-10-14 05:55:21 +08:00
dnl Configure may need to alter any of the *FLAGS variables in order for
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
dnl various checks to work correctly. Save the user's value here so it
dnl can be restored once all configure checks are complete.
2009-10-14 05:55:21 +08:00
saved_user_CFLAGS="$CFLAGS"
saved_user_CXXFLAGS="$CXXFLAGS"
saved_user_FCFLAGS="$FCFLAGS"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
saved_user_LDFLAGS="$LDFLAGS"
saved_user_CPPFLAGS="$CPPFLAGS"
2006-04-27 05:19:21 +08:00
2003-07-31 04:06:56 +08:00
dnl Different compilers may need default libraries. They are specified in
dnl the config/* files, so we put this statement here so that it'll be
dnl set by the code which follows...
dnl
2005-02-17 03:27:27 +08:00
DEFAULT_LIBS=""
2003-07-31 04:06:56 +08:00
2005-03-10 02:48:23 +08:00
dnl Support F9X variable to define Fortran compiler if FC variable is
2007-07-13 06:01:35 +08:00
dnl not used. This should be deprecated in the future.
2005-03-10 02:48:23 +08:00
if test "x" = "x$FC"; then
FC=${F9X}
fi
2003-07-10 02:12:42 +08:00
dnl ----------------------------------------------------------------------
dnl Set prefix default (install directory) to a directory in the build area.
dnl This allows multiple src-dir builds within one host.
AC_PREFIX_DEFAULT([`pwd`/hdf5])
1999-02-11 23:56:15 +08:00
dnl ----------------------------------------------------------------------
dnl Dump all shell variables values.
dnl
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([shell variables initial values])
set >&AS_MESSAGE_LOG_FD
AC_MSG_RESULT([done])
1999-02-11 23:56:15 +08:00
2008-05-15 13:12:00 +08:00
dnl Define all symbol variables used for configure summary.
dnl EXTERNAL_FILTERS equals all external filters. Default none.
dnl MPE: whether MPE option is enabled. Default no.
dnl STATIC_EXEC: whether static-exec is enabled. Default no.
dnl HDF_FORTRAN: whether Fortran is enabled. Default no.
dnl FC: Fortran compiler.
dnl HDF_CXX: whether C++ is enabled. Default no.
dnl CXX: C++ compiler.
dnl HDF5_HL: whether high-level library is enabled. Default is yes.
dnl GPFS: whether gpfs is enabled. Default no.
2010-05-06 01:34:26 +08:00
dnl LARGEFILE: whether largefile support is enabled. Default yes.
2008-05-16 12:00:28 +08:00
dnl INSTRUMENT: whether INSTRUMENT is enabled. No default set here.
2008-05-15 13:12:00 +08:00
dnl CODESTACK: whether CODESTACK is enabled. Default no.
dnl HAVE_DMALLOC: whether system has dmalloc support. Default no.
dnl DIRECT_VFD: whether DIRECT_VFD is enabled. Default no.
dnl THREADSAFE: whether THREADSAFE is enabled. Default no.
dnl STATIC_SHARED: whether static and/or shared libraries are requested.
dnl enable_shared: whether shared lib is enabled.
dnl enable_static: whether static lib is enabled.
dnl UNAME_INFO: System information.
AC_SUBST(EXTERNAL_FILTERS)
AC_SUBST(MPE) MPE=no
AC_SUBST(STATIC_EXEC) STATIC_EXEC=no
AC_SUBST(HDF_FORTRAN) HDF_FORTRAN=no
AC_SUBST(FC) HDF_FORTRAN=no
AC_SUBST(HDF_CXX) HDF_CXX=no
AC_SUBST(CXX) HDF_CXX=no
AC_SUBST(HDF5_HL) HDF5_HL=yes
AC_SUBST(GPFS) GPFS=no
2010-05-06 01:34:26 +08:00
AC_SUBST(LARGEFILE) LARGEFILE=yes
2008-05-16 12:00:28 +08:00
AC_SUBST(INSTRUMENT)
2008-05-15 13:12:00 +08:00
AC_SUBST(CODESTACK) CODESTACK=no
AC_SUBST(HAVE_DMALLOC) HAVE_DMALLOC=no
AC_SUBST(DIRECT_VFD) DIRECT_VFD=no
AC_SUBST(THREADSAFE) THREADSAFE=no
AC_SUBST(STATIC_SHARED)
AC_SUBST(enable_shared)
AC_SUBST(enable_static)
AC_SUBST(UNAME_INFO) UNAME_INFO=`uname -a`
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
2006-05-19 20:22:07 +08:00
dnl Some platforms have broken basename, and/or xargs programs. Check
2003-10-29 01:57:10 +08:00
dnl that it actually does what it's supposed to do. Catch this early
2006-05-19 20:22:07 +08:00
dnl since configure relies upon them heavily and there's no use continuing
2003-10-29 01:57:10 +08:00
dnl if it's broken.
2003-10-14 04:56:27 +08:00
dnl
2005-10-18 05:38:38 +08:00
2006-05-19 20:22:07 +08:00
dnl Avoid depending upon Character Ranges.
dnl These are defined by autoconf.
dnl as_cr_letters='abcdefghijklmnopqrstuvwxyz'
dnl as_cr_LETTERS='ABCDEFGHIJKLMNOPQRSTUVWXYZ'
2003-10-14 04:56:27 +08:00
2003-10-29 01:57:10 +08:00
AC_MSG_CHECKING([if basename works])
BASENAME_TEST="`basename /foo/bar/baz/qux/basename_works`"
if test $BASENAME_TEST != "basename_works"; then
AC_MSG_ERROR([basename program doesn't work])
else
AC_MSG_RESULT([yes])
fi
AC_MSG_CHECKING([if xargs works])
XARGS_TEST="`echo /foo/bar/baz/qux/xargs_works | xargs basename`"
if test $XARGS_TEST != "xargs_works"; then
AC_MSG_ERROR([xargs program doesn't work])
else
AC_MSG_RESULT([yes])
fi
2003-10-14 04:56:27 +08:00
dnl ----------------------------------------------------------------------
1997-08-16 00:35:53 +08:00
dnl Check that the cache file was build on the same host as what we're
dnl running on now.
dnl
2002-07-18 00:10:46 +08:00
AC_CACHE_CHECK([for cached host], [hdf5_cv_host], [hdf5_cv_host="none"]);
1997-08-16 00:35:53 +08:00
if test $hdf5_cv_host = "none"; then
2001-09-28 07:05:16 +08:00
hdf5_cv_host=$host
1997-08-16 00:35:53 +08:00
elif test $hdf5_cv_host != $host; then
2001-09-28 07:05:16 +08:00
echo "The config.cache file was generated on $hdf5_cv_host but"
echo "this is $host. Please remove that file and try again."
AC_MSG_ERROR(config.cache file is invalid)
1997-08-16 00:35:53 +08:00
fi
dnl ----------------------------------------------------------------------
dnl Source any special files that we need. These files normally aren't
dnl present but can be used by the maintainers to fine tune things like
1998-02-14 02:11:35 +08:00
dnl turning on debug or profiling flags for the compiler. The search order
dnl is:
dnl
dnl CPU-VENDOR-OS
dnl VENDOR-OS
dnl CPU-OS
dnl CPU-VENDOR
dnl OS
dnl VENDOR
dnl CPU
1997-08-16 00:35:53 +08:00
dnl
1999-03-03 01:15:35 +08:00
dnl If the `OS' ends with a version number then remove it. For instance,
dnl `freebsd3.1' would become `freebsd'
case $host_os in
2011-04-16 04:10:47 +08:00
aix*)
host_os_novers=aix
2010-09-23 06:52:55 +08:00
;;
2001-09-28 07:05:16 +08:00
freebsd*)
host_os_novers=freebsd
;;
irix5.*)
host_os_novers=irix5.x
;;
irix6.*)
host_os_novers=irix6.x
;;
osf4.*)
host_os_novers=osf4.x
;;
osf5.*)
host_os_novers=osf5.x
;;
solaris2.*)
host_os_novers=solaris2.x
;;
*)
host_os_novers=$host_os
;;
1999-03-03 01:15:35 +08:00
esac
1998-11-13 23:06:06 +08:00
1998-07-09 02:41:04 +08:00
host_config="none"
1999-03-13 02:35:04 +08:00
for f in $host_cpu-$host_vendor-$host_os \
2002-04-23 07:44:41 +08:00
$host_cpu-$host_vendor-$host_os_novers \
$host_vendor-$host_os \
1999-03-03 01:15:35 +08:00
$host_vendor-$host_os_novers \
2002-04-23 07:44:41 +08:00
$host_cpu-$host_os \
1999-03-03 01:15:35 +08:00
$host_cpu-$host_os_novers \
1998-02-14 02:11:35 +08:00
$host_cpu-$host_vendor \
2002-04-23 07:44:41 +08:00
$host_os \
$host_os_novers \
$host_vendor \
$host_cpu ; do
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for config $f])
2001-08-02 05:00:25 +08:00
if test -f "$srcdir/config/$f"; then
host_config=$srcdir/config/$f
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([found])
2001-08-02 05:00:25 +08:00
break
fi
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([no])
1997-08-16 00:35:53 +08:00
done
2003-01-24 03:19:02 +08:00
if test "X$host_config" != "Xnone"; then
2001-09-28 07:05:16 +08:00
CC_BASENAME="`echo $CC | cut -f1 -d' ' | xargs basename 2>/dev/null`"
2001-08-02 05:00:25 +08:00
. $host_config
1997-08-16 00:35:53 +08:00
fi
2000-10-04 06:28:31 +08:00
dnl Source any special site-specific file
hname="`hostname`"
while test -n "$hname"; do
2001-08-02 05:00:25 +08:00
file=$srcdir/config/site-specific/host-$hname
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for config $file])
2001-08-02 05:00:25 +08:00
if test -f "$file"; then
. $file
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([found])
2001-08-02 05:00:25 +08:00
break
fi
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([no])
2001-08-02 05:00:25 +08:00
hname_tmp=$hname
hname="`echo $hname | cut -d. -f2-99`"
test "$hname_tmp" = "$hname" && break
2000-10-04 06:28:31 +08:00
done
2009-10-19 11:05:25 +08:00
dnl ----------------------------------------------------------------------
dnl Some built-in configure checks can only see CFLAGS (not AM_CFLAGS), so
dnl we need to add this in so configure works as intended. We will need to
dnl reset this value at the end of configure, to preserve the user's settings.
CFLAGS="${AM_CFLAGS} ${CFLAGS}"
FCFLAGS="${AM_FCFLAGS} ${FCFLAGS}"
CXXFLAGS="${AM_CXXFLAGS} ${CXXFLAGS}"
CPPFLAGS="${AM_CPPFLAGS} ${CPPFLAGS}"
LDFLAGS="${AM_LDFLAGS} ${LDFLAGS}"
2005-03-15 04:35:55 +08:00
dnl ----------------------------------------------------------------------
dnl Enable dependency tracking unless the configure options or a
dnl site-specific file told us not to. This prevents configure from
dnl silently disabling dependencies for some compilers.
dnl
if test -z "${enable_dependency_tracking}"; then
enable_dependency_tracking="yes"
fi
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
dnl Check for programs.
dnl
AC_PROG_CC
2001-09-28 07:05:16 +08:00
CC_BASENAME="`echo $CC | cut -f1 -d' ' | xargs basename 2>/dev/null`"
2000-09-08 04:57:34 +08:00
2010-11-18 02:07:51 +08:00
dnl ----------------------------------------------------------------------------
dnl Configure disallows unsupported combinations of options. However, users
dnl may want to override and build with unsupported combinations for their
dnl own use. They can use the --enable-unsupported configure flag, which
dnl ignores any errors from configure due to incompatible flags.
AC_MSG_CHECKING([if unsupported combinations of configure options are allowed])
AC_ARG_ENABLE([unsupported],
[AC_HELP_STRING([--enable-unsupported],
[Allow unsupported combinations of configure options])],
[ALLOW_UNSUPPORTED=$enableval])
case "X-$ALLOW_UNSUPPORTED" in
X-|X-no)
AC_MSG_RESULT([no])
;;
X-yes)
AC_MSG_RESULT([yes])
;;
*)
;;
esac
2000-09-08 04:57:34 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they would like the Fortran interface compiled
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([HDF5_INTERFACES]) HDF5_INTERFACES=""
AC_MSG_CHECKING([if Fortran interface enabled])
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([fortran],
[AC_HELP_STRING([--enable-fortran],
[Compile the Fortran interface [default=no]])],
[HDF_FORTRAN=$enableval])
2000-09-08 04:57:34 +08:00
2008-09-02 23:04:59 +08:00
HAVE_SIZEOF="no"
FORTRAN_DEFAULT_REALisDBLE="no"
2000-09-15 06:19:24 +08:00
if test "X$HDF_FORTRAN" = "Xyes"; then
2001-08-02 05:00:25 +08:00
echo "yes"
2003-07-29 05:38:04 +08:00
HDF5_INTERFACES="$HDF5_INTERFACES fortran"
dnl --------------------------------------------------------------------
dnl HDF5 integer variables for the H5fortran_types.f90 file.
dnl
AC_SUBST([R_LARGE])
AC_SUBST([R_INTEGER])
2005-01-11 06:56:20 +08:00
AC_SUBST([HADDR_T])
2003-07-29 05:38:04 +08:00
AC_SUBST([HSIZE_T])
AC_SUBST([HSSIZE_T])
AC_SUBST([HID_T])
AC_SUBST([SIZE_T])
AC_SUBST([OBJECT_NAMELEN_DEFAULT_F])
dnl --------------------------------------------------------------------
dnl General Fortran flags
dnl
2009-10-19 11:05:25 +08:00
AM_FCFLAGS="${AM_FCFLAGS} ${FFLAGS}"
FCFLAGS="${FCFLAGS} ${FFLAGS}"
2005-03-10 02:48:23 +08:00
2005-03-11 04:52:34 +08:00
dnl --------------------------------------------------------------------
dnl Fortran source extention
dnl
2006-11-11 03:56:04 +08:00
AC_FC_SRCEXT([f90])
2005-03-11 04:52:34 +08:00
2003-07-29 05:38:04 +08:00
AC_SUBST([F9XSUFFIXFLAG])
AC_SUBST([FSEARCH_DIRS])
dnl --------------------------------------------------------------------
dnl Check for a Fortran 9X compiler and how to include modules.
dnl
2005-07-19 07:30:38 +08:00
AC_PROG_FC([f90 pgf90 slf90 f95 g95 xlf95 efc ifort ftn],)
2003-07-29 05:38:04 +08:00
AC_F9X_MODS
[svn-r10158] Purpose:
Automake version upgrade
Description:
Upgraded automake version from 1.6.2 to 1.9.5.
Changed bin/reconfigure script to use automake 1.9.5.
Changed configure.in and Makefiles to use new FCFLAGS and FC variables
instead of FFLAGS and F9X.
Automake and configure should now do the lion's share of the work
supporting Fortran 9X; macros in acsite.m4 are now mostly unused (will
be cleaned later).
Altered how configure handles pmake; now root-level Makefile.in is
processed by bin/reconfigure to have a .MAKEFLAGS target, since
automake no longer allows us to define unused variables.
Configure now always checks for C++ compiler even if it is not
used, since automake thinks this is the Right Thing To Do and
will break otherwise.
Platforms tested:
Sol, copper, heping, mir, sleipnir, eirene, pommier, kelgia, modi4.
2005-03-08 01:57:27 +08:00
dnl It seems that libtool (as of Libtool 1.5.14) is trying to
dnl configure itself for Fortran 77.
dnl Tell it that our F77 compiler is $FC (actually a F9X compiler)
F77=$FC
2005-02-01 11:17:02 +08:00
2003-07-29 05:38:04 +08:00
dnl Change to the Fortran 90 language
2005-03-11 04:52:34 +08:00
AC_LANG_PUSH(Fortran)
2003-07-29 05:38:04 +08:00
2005-04-05 05:17:51 +08:00
dnl --------------------------------------------------------------------
dnl Define wrappers for the C compiler to use Fortran function names
dnl
AC_FC_WRAPPERS
2003-07-29 05:38:04 +08:00
dnl --------------------------------------------------------------------
dnl See if the compiler will support the "-I." option
dnl
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
dnl AM_FCFLAGS_saved=$AM_FCFLAGS
dnl AM_FCFLAGS="${AM_FCFLAGS} -I."
2005-04-05 05:17:51 +08:00
dnl AC_MSG_CHECKING(if compiler supports -I. option)
dnl AC_TRY_FCOMPILE([
dnl program conftest
dnl end
dnl ], AC_MSG_RESULT(yes),
dnl AC_MSG_RESULT(no)
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
dnl AM_FCFLAGS="$AM_FCFLAGS_saved")
2003-07-29 05:38:04 +08:00
2008-09-01 23:25:02 +08:00
dnl --------------------------------------------------------------------
dnl See if the fortran compiler supports the intrinsic function "SIZEOF"
AC_MSG_CHECKING([if Fortran compiler supports intrinsic SIZEOF])
2008-09-03 23:40:38 +08:00
AC_TRY_RUN([
PROGRAM main
i = sizeof(x)
END PROGRAM
2008-09-01 23:25:02 +08:00
], [AC_MSG_RESULT(yes)
HAVE_SIZEOF="yes"],
AC_MSG_RESULT(no))
2008-09-02 23:04:59 +08:00
2008-09-01 23:25:02 +08:00
dnl Check to see if -r8 was specified to determine if we need to
dnl compile the DOUBLE PRECISION interfaces.
2008-08-27 00:05:06 +08:00
AC_MSG_CHECKING([if Fortran default REAL is DOUBLE PRECISION])
AC_TRY_RUN([
MODULE type_mod
INTERFACE h5t
MODULE PROCEDURE h5t_real
MODULE PROCEDURE h5t_dble
END INTERFACE
CONTAINS
SUBROUTINE h5t_real(r)
REAL :: r
END SUBROUTINE h5t_real
SUBROUTINE h5t_dble(d)
DOUBLE PRECISION :: d
END SUBROUTINE h5t_dble
END MODULE type_mod
PROGRAM main
USE type_mod
REAL :: r
DOUBLE PRECISION :: d
CALL h5t(r)
CALL h5t(d)
END PROGRAM main
],
AC_MSG_RESULT(no),
[AC_MSG_RESULT(yes)
FORTRAN_DEFAULT_REALisDBLE="yes"])
2005-03-11 04:52:34 +08:00
dnl Change back to the C language
AC_LANG_POP(Fortran)
2000-09-08 04:57:34 +08:00
else
2001-08-02 05:00:25 +08:00
echo "no"
2009-10-02 01:05:38 +08:00
FC="no"
2000-09-08 04:57:34 +08:00
fi
2008-09-02 23:04:59 +08:00
AM_CONDITIONAL([FORTRAN_HAVE_SIZEOF], [test "X$HAVE_SIZEOF" = "Xyes"])
2008-08-27 00:05:06 +08:00
AM_CONDITIONAL([FORTRAN_DEFAULT_REALisDBLE_F], [test "X$FORTRAN_DEFAULT_REALisDBLE" = "Xyes"])
2000-09-08 04:57:34 +08:00
dnl ----------------------------------------------------------------------
2000-11-14 05:25:12 +08:00
dnl Check if they would like the C++ interface compiled
2000-09-08 04:57:34 +08:00
dnl
[svn-r10158] Purpose:
Automake version upgrade
Description:
Upgraded automake version from 1.6.2 to 1.9.5.
Changed bin/reconfigure script to use automake 1.9.5.
Changed configure.in and Makefiles to use new FCFLAGS and FC variables
instead of FFLAGS and F9X.
Automake and configure should now do the lion's share of the work
supporting Fortran 9X; macros in acsite.m4 are now mostly unused (will
be cleaned later).
Altered how configure handles pmake; now root-level Makefile.in is
processed by bin/reconfigure to have a .MAKEFLAGS target, since
automake no longer allows us to define unused variables.
Configure now always checks for C++ compiler even if it is not
used, since automake thinks this is the Right Thing To Do and
will break otherwise.
Platforms tested:
Sol, copper, heping, mir, sleipnir, eirene, pommier, kelgia, modi4.
2005-03-08 01:57:27 +08:00
dnl We need to check for a C++ compiler unconditionally, since
dnl AC_PROG_CXX defines some macros that Automake 1.9.x uses and will
dnl miss even if c++ is not enabled.
AC_PROG_CXX
AC_PROG_CXXCPP dnl this is checked for when AC_HEADER_STDC is done
2006-06-15 04:16:27 +08:00
AC_MSG_CHECKING([if c++ interface enabled])
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([cxx],
[AC_HELP_STRING([--enable-cxx],
[Compile the C++ interface [default=no]])],
[HDF_CXX=$enableval])
2000-09-08 04:57:34 +08:00
2000-09-15 06:19:24 +08:00
if test "X$HDF_CXX" = "Xyes"; then
2001-08-02 05:00:25 +08:00
echo "yes"
2003-07-29 05:38:04 +08:00
HDF5_INTERFACES="$HDF5_INTERFACES c++"
dnl Change to the C++ language
2005-03-11 04:52:34 +08:00
AC_LANG_PUSH(C++)
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([if $CXX needs old style header files in includes])
AC_TRY_RUN([
#include <iostream>
int main(void) { return 0; }
], [
echo no
], [
echo yes
2009-10-19 11:05:25 +08:00
CXXFLAGS="${CXXFLAGS} -DOLD_HEADER_FILENAME"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CXXFLAGS="${AM_CXXFLAGS} -DOLD_HEADER_FILENAME"
2003-07-29 05:38:04 +08:00
])
AC_MSG_CHECKING([if $CXX can handle namespaces])
AC_TRY_RUN([
namespace H5 {
int fnord;
}
int main(void) {
using namespace H5;
fnord = 37;
return 0;
}
], [
echo yes
], [
echo no
2009-10-19 11:05:25 +08:00
CXXFLAGS="${CXXFLAGS} -DH5_NO_NAMESPACE"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CXXFLAGS="${AM_CXXFLAGS} -DH5_NO_NAMESPACE"
2003-07-29 05:38:04 +08:00
])
AC_MSG_CHECKING([if $CXX supports std])
AC_TRY_RUN([
#include <string>
using namespace std;
int main(void) {
string myString("testing namespace std");
return 0;
}
], [
echo yes
], [
echo no
2009-10-19 11:05:25 +08:00
CXXFLAGS="${CXXFLAGS} -DH5_NO_STD"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CXXFLAGS="${AM_CXXFLAGS} -DH5_NO_STD"
2003-07-29 05:38:04 +08:00
])
AC_MSG_CHECKING([if $CXX supports bool types])
AC_TRY_RUN([
int main(void) {
bool flag;
return 0;
}
], [
echo yes
], [
echo no
2009-10-19 11:05:25 +08:00
CXXFLAGS="${CXXFLAGS} -DBOOL_NOTDEFINED"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CXXFLAGS="${AM_CXXFLAGS} -DBOOL_NOTDEFINED"
2003-07-29 05:38:04 +08:00
])
2007-06-29 04:36:08 +08:00
AC_MSG_CHECKING([if $CXX has offsetof extension])
AC_TRY_COMPILE([
#include <stdio.h>
#include <stddef.h>
],[
struct index_st
{
unsigned char type;
unsigned char num;
unsigned int len;
};
typedef struct index_st index_t;
int x,y;
x = offsetof(struct index_st, len);
y = offsetof(index_t, num)
],
2007-06-30 04:52:56 +08:00
AC_DEFINE([CXX_HAVE_OFFSETOF], [1],
2007-06-29 04:36:08 +08:00
[Define if C++ compiler recognizes offsetof])
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([if $CXX can handle static cast])
AC_TRY_RUN([
int main(void) {
float test_float;
int test_int;
test_float = 37.0;
test_int = static_cast <int> (test_float);
return 0;
}
], [
echo yes
], [
echo no
2009-10-19 11:05:25 +08:00
CXXFLAGS="${CXXFLAGS} -DNO_STATIC_CAST"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CXXFLAGS="${AM_CXXFLAGS} -DNO_STATIC_CAST"
2003-07-29 05:38:04 +08:00
])
2000-09-08 04:57:34 +08:00
else
2001-08-02 05:00:25 +08:00
echo "no"
2009-10-02 01:05:38 +08:00
CXX="no"
2000-09-08 04:57:34 +08:00
fi
2006-06-15 04:16:27 +08:00
dnl Change back to the C language
AC_LANG_POP(C++)
2001-08-02 05:00:25 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they have Perl installed on their system. We only need Perl
dnl if they're using a GNU compiler.
dnl
2002-07-18 00:10:46 +08:00
AC_SUBST([PERL]) PERL=""
2001-08-02 05:00:25 +08:00
if test "X$GCC" = "Xyes"; then
2002-07-18 00:10:46 +08:00
AC_CHECK_PROGS([PERL], [perl],, [$PATH])
2001-08-02 05:00:25 +08:00
fi
[svn-r10158] Purpose:
Automake version upgrade
Description:
Upgraded automake version from 1.6.2 to 1.9.5.
Changed bin/reconfigure script to use automake 1.9.5.
Changed configure.in and Makefiles to use new FCFLAGS and FC variables
instead of FFLAGS and F9X.
Automake and configure should now do the lion's share of the work
supporting Fortran 9X; macros in acsite.m4 are now mostly unused (will
be cleaned later).
Altered how configure handles pmake; now root-level Makefile.in is
processed by bin/reconfigure to have a .MAKEFLAGS target, since
automake no longer allows us to define unused variables.
Configure now always checks for C++ compiler even if it is not
used, since automake thinks this is the Right Thing To Do and
will break otherwise.
Platforms tested:
Sol, copper, heping, mir, sleipnir, eirene, pommier, kelgia, modi4.
2005-03-08 01:57:27 +08:00
2000-09-26 01:35:08 +08:00
dnl ----------------------------------------------------------------------
dnl Check which archiving tool to use. This needs to be done before
2005-02-01 11:17:02 +08:00
dnl the AM_PROG_LIBTOOL macro.
2000-09-26 01:35:08 +08:00
dnl
if test -z "$AR"; then
2002-07-18 00:10:46 +08:00
AC_CHECK_PROGS([AR], [ar xar], [:], [$PATH])
1998-03-06 13:34:36 +08:00
fi
2002-07-18 00:10:46 +08:00
AC_SUBST([AR])
1998-03-06 13:34:36 +08:00
2000-09-26 01:35:08 +08:00
dnl Export the AR macro so that it will be placed in the libtool file
dnl correctly.
export AR
2004-12-29 22:26:20 +08:00
AC_PROG_MAKE_SET
AC_PROG_INSTALL
[svn-r11411] Purpose:
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
2005-09-14 06:30:33 +08:00
2006-06-15 04:16:27 +08:00
dnl ----------------------------------------------------------------------
2006-10-04 06:41:34 +08:00
dnl Check that the tr utility is working properly.
AC_PATH_PROG(TR, tr)
TR_TEST=`echo Test | ${TR} ${as_cr_letters}"," ${as_cr_LETTERS}" "`
if test "X${TR_TEST}" != "XTEST"; then
AC_MSG_ERROR([tr program doesn't work])
fi
2007-01-24 01:29:45 +08:00
dnl ----------------------------------------------------------------------
dnl Check that time can be used with srcdir. This is okay on most systems,
dnl but seems to cause problems on Cygwin.
dnl The solution on Cygwin is not to record execution time for tests.
AC_MSG_CHECKING([if srcdir= and time commands work together])
AC_SUBST([TIME])
TIME=time
2007-01-25 23:30:53 +08:00
TIME_TEST=`foo="bar" ${TIME} echo 'baz' 2> /dev/null | grep baz`
2007-01-24 01:29:45 +08:00
if test "X${TIME_TEST}" = "Xbaz"; then
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
TIME=
fi
2006-06-15 04:16:27 +08:00
dnl The following variables are used to distinguish between building a
dnl serial and parallel library.
dnl
dnl HAVE_PARALLEL -- defined in H5config.h if we are building
dnl a parallel library even if configure wasn't
dnl able to find some header file or library that
dnl might be required. This is defined if the
dnl compiler looks like a parallel compiler (e.g.,
dnl mpicc or mpcc) or if the user explicitly states
dnl that a parallel library is being built by supplying
dnl the `--enable-parallel' configure switch.
dnl
dnl PARALLEL -- This variable is set to a non-null value if
dnl configure thinks we're compiling a parallel
dnl version of the library.
dnl
dnl RUNSERIAL -- This is a command which will be prepended to
dnl the executable name to run the executable using
dnl a single process. For serial versions of the
dnl library this will normally be empty. For parallel
2010-06-12 07:37:33 +08:00
dnl versions it might be something like `mpiexec -n 1'.
2006-06-15 04:16:27 +08:00
dnl The value of this variable is substituted in *.in
dnl files.
dnl
dnl RUNPARALLEL -- This is a command which will be prepended to
dnl the executable name to run the executable on
dnl multiple processors. For the serial library the
dnl value will normally be the empty string. For
dnl parallel library it should be something like
2010-10-29 01:19:30 +08:00
dnl "mpiexec -n \$\${NPROCS:=6}" where NPROCS will
dnl eventually contain the number of processors on which
dnl to run the executable (the double dollarsigns are to
2006-06-15 04:16:27 +08:00
dnl protect the expansion until make executes the
dnl command). The value of this variable is
dnl substituted in *.in files.
dnl
AC_SUBST([PARALLEL])
AC_SUBST([RUNSERIAL])
AC_SUBST([RUNPARALLEL])
AC_SUBST([TESTPARALLEL])
dnl ----------------------------------------------------------------------
dnl If the compiler is obviously a parallel compiler then we're building
dnl a parallel version of hdf5 and should define HAVE_PARALLEL. Furthermore,
dnl the name of the compiler might tell us how to run the resulting
2010-06-12 07:37:33 +08:00
dnl executable. For `mpicc' the executable should be run with `mpiexec' from
2006-06-15 04:16:27 +08:00
dnl the same directory as mpicc if it exists.
dnl
case "$CC_BASENAME" in
mpicc)
2010-06-12 07:37:33 +08:00
dnl The mpich compiler. Use mpiexec from the same directory if it
2006-06-15 04:16:27 +08:00
dnl exists.
PARALLEL=mpicc
2010-06-12 07:37:33 +08:00
AC_MSG_CHECKING([for mpiexec])
2006-06-15 04:16:27 +08:00
dnl Find the path where mpicc is located.
cmd="`echo $CC | cut -f1 -d' '`"
if (echo $cmd | grep / >/dev/null); then
path="`echo $cmd | sed 's/\(.*\)\/.*$/\1/'`"
else
2006-10-04 06:41:34 +08:00
for path in `echo $PATH | ${TR} ":" " "`; do
2006-06-15 04:16:27 +08:00
if test -x $path/$cmd; then
break
fi
done
fi
2010-06-12 07:37:33 +08:00
dnl Is there an mpiexec at that path?
if test -x $path/mpiexec; then
AC_MSG_RESULT([$path/mpiexec])
2006-06-15 04:16:27 +08:00
RUNSERIAL="${RUNSERIAL:-none}"
if test -z "$RUNPARALLEL"; then
2010-10-29 01:19:30 +08:00
RUNPARALLEL="$path/mpiexec -n \$\${NPROCS:=6}"
2006-06-15 04:16:27 +08:00
fi
else
AC_MSG_RESULT([none])
fi
;;
mpcc|mpcc_r)
dnl The IBM compiler
PARALLEL="$CC_BASENAME"
;;
*)
dnl Probably not a parallel compiler, but if `--enable-parallel'
dnl is defined below then we're still building a parallel hdf5.
;;
esac
dnl ----------------------------------------------------------------------
dnl If the Fortran compiler is obviously a parallel compiler then we're
dnl building a parallel version of hdf5 and should define HAVE_PARALLEL.
dnl Furthermore, the name of the compiler might tell us how to run the
dnl resulting executable. For `mpif90' the executable should be run with
2010-06-12 07:37:33 +08:00
dnl `mpiexec' from the same directory as mpif90 if it exists.
2006-06-15 04:16:27 +08:00
dnl
if test "X$HDF_FORTRAN" = "Xyes"; then
dnl Change to the Fortran 90 language
AC_LANG_PUSH(Fortran)
case "$FC" in
*mpif90*)
2010-06-12 07:37:33 +08:00
dnl The Fortran mpich compiler. Use mpiexec from the same directory
2006-06-15 04:16:27 +08:00
dnl if it exists.
PARALLEL=mpif90
2010-06-12 07:37:33 +08:00
AC_MSG_CHECKING([for mpiexec])
2006-06-15 04:16:27 +08:00
dnl Find the path where mpif90 is located.
cmd=`echo $FC |cut -f1 -d' '`
if (echo $cmd |grep / >/dev/null); then
path="`echo $cmd |sed 's/\(.*\)\/.*$/\1/'`"
else
2006-10-04 06:41:34 +08:00
for path in `echo $PATH | ${TR} ":" " "`; do
2006-06-15 04:16:27 +08:00
if test -x $path/$cmd; then
break;
fi
done
fi
2010-06-12 07:37:33 +08:00
dnl Is there an mpiexec at that path?
if test -x $path/mpiexec; then
AC_MSG_RESULT([$path/mpiexec])
2006-06-15 04:16:27 +08:00
RUNSERIAL="${RUNSERIAL:-none}"
if test -z "$RUNPARALLEL"; then
2010-10-29 01:19:30 +08:00
RUNPARALLEL="$path/mpiexec -n \$\${NPROCS:=6}"
2006-06-15 04:16:27 +08:00
fi
else
AC_MSG_RESULT([none])
fi
;;
*mpxlf* | *mpxlf_r* | *mpxlf90* | *mpxlf90_r* | *mpxlf95* | *mpxlf95_r*)
dnl The IBM compiler
PARALLEL="$FC"
;;
*)
dnl Probably not a parallel compiler, but if `--enable-parallel'
dnl is defined below then we're still building a parallel hdf5.
;;
esac
dnl Change to the C language
AC_LANG_POP(Fortran)
fi
dnl -----------------------------------------------------------------------------
dnl If shared libraries are being used with parallel, disable them, unless the
dnl user explicity enables them via the '--enable-shared' option.
if test "X${enable_shared}" = "X" -a "X${enable_parallel}" = "Xyes"; then
echo ' shared libraries disabled in parallel'
enable_shared="no"
elif test "X${enable_shared}" = "Xyes" -a "X${enable_parallel}" = "Xyes"; then
echo ' shared libraries explicitly enabled by user'
elif test "X${enable_shared}" = "X" -a "X${PARALLEL}" != "X"; then
echo ' shared libraries disabled when a parallel compiler is being used'
enable_shared="no"
elif test "X${enable_shared}" = "Xyes" -a "X${PARALLEL}" != "X"; then
echo ' shared libraries explicitly enabled by user'
fi
2005-09-16 05:30:06 +08:00
dnl ----------------------------------------------------------------------
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl Fortran libraries are not currently supported on Mac. Disable them.
dnl (this is overridable with --enable-unsupported).
dnl
AC_SUBST([H5_FORTRAN_SHARED])
H5_FORTRAN_SHARED="no"
if test "X${HDF_FORTRAN}" = "Xyes" && test "X${enable_shared}" != "Xno"; then
AC_MSG_CHECKING([if shared Fortran libraries are supported])
2005-09-16 05:30:06 +08:00
H5_FORTRAN_SHARED="yes"
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl Disable fortran shared libraries on Mac. (MAM - 03/30/11)
case "`uname`" in
Darwin*)
2005-09-16 05:30:06 +08:00
H5_FORTRAN_SHARED="no"
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
CHECK_WARN="Shared Fortran libraries not currently supported on Mac."
;;
esac
2005-09-16 05:30:06 +08:00
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl Report results of check(s)
if test "X${H5_FORTRAN_SHARED}" = "Xno"; then
AC_MSG_RESULT([no])
AC_MSG_WARN([$CHECK_WARN])
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
AC_MSG_WARN([Disabling shared Fortran libraries.])
AC_MSG_WARN([To override this behavior, please use --enable-unsupported configure option.])
if test "X${enable_static}" = "Xno"; then
AC_MSG_ERROR([both static and shared Fortran libraries are disabled])
fi
2008-08-27 04:10:57 +08:00
else
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
AC_MSG_WARN([Allowing unsupported Fortran shared libraries due to use of --enable-unsupported flag])
H5_FORTRAN_SHARED="yes"
2008-08-27 04:10:57 +08:00
fi
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
else
AC_MSG_RESULT([yes])
2005-09-16 05:30:06 +08:00
fi
[svn-r11411] Purpose:
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
2005-09-14 06:30:33 +08:00
fi
2005-09-16 05:30:06 +08:00
AM_CONDITIONAL([FORTRAN_SHARED_CONDITIONAL], [test "X$H5_FORTRAN_SHARED" = "Xyes"])
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl ----------------------------------------------------------------------
dnl Disable C++ shared libraries if +DD64 flag is detected.
dnl
AC_SUBST([H5_CXX_SHARED])
2006-11-17 04:44:35 +08:00
H5_CXX_SHARED="no"
2005-09-17 05:16:11 +08:00
if test "X${HDF_CXX}" = "Xyes" && test "X${enable_shared}" != "Xno"; then
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
AC_MSG_CHECKING([if shared C++ libraries are supported])
2005-09-17 05:16:11 +08:00
H5_CXX_SHARED="yes"
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl Disable C++ shared libraries if DD64 flag is being used.
2006-12-20 23:22:54 +08:00
if (echo dummy ${CXX} ${CXXLD} ${CFLAGS} ${CXXFLAGS} ${LDFLAGS} | grep 'DD64') > /dev/null; then
H5_CXX_SHARED="no"
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
CHECK_WARN="Shared C++ libraries not currently supported with +DD64 flag."
2006-12-20 23:22:54 +08:00
fi
[svn-r20475] Purpose:
- Revise shared Fortran library disabling scenarios in configure
- Improve configure output summary
Description:
Shared Fortran libraries are not supported on Mac, but were being
disabled by configure in a way that also forced the C libraries
to be static-only. This has been fixed, so now only shared Fortran
is disabled while shared C can remain.
This prompted two additional changes:
1. While working on the check that addresses whether or not
shared Fortran libraries are allowed, removed old and no
longer needed check(s) that disable shared Fortran
libraries with HP, Intel 8, PGI, and Absoft compilers.
(Essentially, Mac is the only situation in which Fortran
shared are disabled by configure.)
2. Having two different states of libraries (i.e. shared C
library with static-only Fortran library) was not apparent
in the configure summary, which labeled all libraries as
either shared and/or static. I've added lines to both the
C++ and Fortran output sections to list shared/static-ness
of these libraries specifically.
Additionally, I've made sure that the new --enable-unsupported
configure option correctly overrides configure if it tries to
disable a shared library.
Tested:
jam, fred, & h5committest
2011-04-12 05:00:38 +08:00
dnl Report results of check(s)
if test "X${H5_CXX_SHARED}" = "Xno"; then
AC_MSG_RESULT([no])
AC_MSG_WARN([$CHECK_WARN])
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
AC_MSG_WARN([Disabling shared C++ libraries.])
AC_MSG_WARN([To override this behavior, please use --enable-unsupported configure option.])
if test "X${enable_static}" = "Xno"; then
AC_MSG_ERROR([both static and shared C++ libraries are disabled])
fi
else
AC_MSG_WARN([Allowing unsupported C++ shared librares due to use of --enable-unsupported flag])
fi
else
AC_MSG_RESULT([yes])
2005-09-17 05:16:11 +08:00
fi
fi
AM_CONDITIONAL([CXX_SHARED_CONDITIONAL], [test "X$H5_CXX_SHARED" = "Xyes"])
[svn-r11411] Purpose:
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
2005-09-14 06:30:33 +08:00
dnl ----------------------------------------------------------------------
dnl pgcc version 6.0x have optimization (-O, -O2 or -O3) problem. Detect
dnl these versions and add option "-Mx,28,0x8" to the compiler to avoid
dnl the problem if optimization is enabled.
dnl
if (${CC-cc} -V 2>&1 | grep '^pgcc 6.0') > /dev/null && test "X$enable_production" = "Xyes"; then
echo 'adding compiler flag to avoid optimization problem in pgcc'
CC="${CC-cc} -Mx,28,0x8"
fi
dnl ----------------------------------------------------------------------
2010-11-18 02:07:51 +08:00
dnl Shared libraries are not currently supported under Cygwin, so configure
dnl disables them unless --enable-unsupported has been supplied by the user.
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
case "`uname`" in
CYGWIN*)
if test "X${enable_shared}" = "Xyes"; then
echo ' warning: shared libraries are not supported on Cygwin!'
echo ' disabling shared libraries'
echo ' use --enable-unsupported to override this warning and keep shared libraries enabled'
fi
enable_shared="no"
;;
esac
fi
[svn-r11411] Purpose:
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
2005-09-14 06:30:33 +08:00
2011-04-15 05:21:59 +08:00
dnl ----------------------------------------------------------------------
dnl Windows won't create DLLs without the following macro.
dnl
AC_LIBTOOL_WIN32_DLL
[svn-r11411] Purpose:
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
2005-09-14 06:30:33 +08:00
dnl ----------------------------------------------------------------------
dnl Create libtool. If shared/static libraries are going to be enabled
dnl or disabled, it should happen before these macros.
2004-12-29 22:26:20 +08:00
AC_LIBTOOL_DLOPEN
2005-02-01 11:17:02 +08:00
AM_PROG_LIBTOOL
2004-12-29 22:26:20 +08:00
2005-03-01 04:08:54 +08:00
dnl Post processing to patch up some deficiencies in libtool (as of
dnl Libtool 1.5.14)
2003-01-15 23:57:52 +08:00
case $host_os in
2005-03-01 04:08:54 +08:00
linux* | freebsd* )
2003-01-17 13:21:57 +08:00
# If gcc is not used, need to set $wl to use "-Wl,"
if $CC -v 2>&1 | grep '^gcc' > /dev/null ; then
: using gcc
else
2003-01-24 03:19:02 +08:00
echo 'fixing $wl in' $ofile
2003-01-15 23:57:52 +08:00
ed - $ofile <<EOF 2> /dev/null
/^wl=""/s//wl="-Wl,"/
w
q
EOF
fi
;;
esac
2009-09-01 00:21:56 +08:00
dnl ----------------------------------------------------------------------
dnl Check if we should install only statically linked executables.
dnl This check needs to occur after libtool is initialized because
dnl we check a libtool cache value and may issue a warning based
dnl on its result.
AC_MSG_CHECKING([if we should install only statically linked executables])
AC_ARG_ENABLE([static_exec],
[AC_HELP_STRING([--enable-static-exec],
[Install only statically linked executables
[default=no]])],
[STATIC_EXEC=$enableval])
if test "X$STATIC_EXEC" = "Xyes"; then
echo "yes"
dnl Issue a warning if -static flag is not supported.
if test "X$lt_cv_prog_compiler_static_works" = "Xno"; then
echo " warning: -static flag not supported on this system; executable won't statically link shared system libraries."
fi
LT_STATIC_EXEC="-all-static"
else
echo "no"
LT_STATIC_EXEC=""
fi
AC_SUBST([LT_STATIC_EXEC])
2001-02-13 06:52:00 +08:00
dnl Fix up the INSTALL macro if it's a relative path. We want the
dnl full-path to the binary instead.
case "$INSTALL" in
2001-08-02 05:00:25 +08:00
*install-sh*)
INSTALL='\${top_srcdir}/bin/install-sh -c'
;;
2001-02-13 06:52:00 +08:00
esac
2008-05-16 05:08:23 +08:00
dnl ----------------------------------------------------------------------
dnl Some users have reported problems with libtool's use of '-Wl,-rpath' to
dnl link shared libraries in nondefulat directories. Allow users to
dnl disable embedding the rpath information in the executables and to
dnl instead solely rely on the information in LD_LIBRARY_PATH.
AC_MSG_CHECKING([if -Wl,-rpath should be used to link shared libs in nondefault directories])
AC_ARG_ENABLE([sharedlib-rpath],
[AC_HELP_STRING([--disable-sharedlib-rpath],
[Disable use of the '=Wl,-rpath' linker option])],
[RPATH=$enableval])
case "X-$RPATH" in
X-no)
AC_MSG_RESULT([no])
2008-05-31 05:47:44 +08:00
runpath_var=
hardcode_libdir_flag_spec=
hardcode_libdir_flag_spec_ld=
2008-05-16 05:08:23 +08:00
;;
X-|X-yes)
AC_MSG_RESULT([yes])
;;
*)
AC_MSG_RESULT([error])
AC_MSG_ERROR([\'$enableval\' is not a valid rpath type])
;;
esac
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([make])
1999-11-01 23:21:16 +08:00
2000-09-15 06:19:24 +08:00
dnl ----------------------------------------------------------------------
1999-11-01 23:21:16 +08:00
dnl Sometimes makes think the `.PATH:' appearing before the first rule
dnl with an action should override the `all' default target. So we have
dnl to decide what the proper syntax is.
dnl
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([how make searches directories])
1999-11-01 23:21:16 +08:00
while true; do #for break
2001-09-28 07:05:16 +08:00
# The most common method is `VPATH=DIR1 DIR2 ...'
cat >maketest <<EOF
1999-11-01 23:21:16 +08:00
VPATH=$srcdir/config $srcdir/src $srcdir/bin
.c.o:
cp $< H5.o
foo: H5.o
2007-02-07 02:50:04 +08:00
rm -f H5.o
1999-11-01 23:21:16 +08:00
@echo works
EOF
2000-09-15 06:19:24 +08:00
2001-09-28 07:05:16 +08:00
if (${MAKE-make} -f maketest foo) >/dev/null 2>&1; then
SEARCH_RULE='VPATH='
SEARCH_SEP=' '
AC_MSG_RESULT([VPATH=DIR1 DIR2 ...])
break
fi
dnl The second most common method is like above except with the
dnl directories separated by colons.
cat >maketest <<EOF
1999-11-01 23:21:16 +08:00
VPATH=$srcdir/config:$srcdir/src:$srcdir/bin
.c.o:
cp $< H5.o
foo: H5.o
2007-02-07 02:50:04 +08:00
rm -f H5.o
1999-11-01 23:21:16 +08:00
@echo works
EOF
2000-09-15 06:19:24 +08:00
2001-09-28 07:05:16 +08:00
if (${MAKE-make} -f maketest foo) >/dev/null 2>&1; then
SEARCH_RULE='VPATH='
SEARCH_SEP=':'
AC_MSG_RESULT([VPATH=DIR1:DIR2:...])
break
fi
dnl pmake uses the construct `.PATH: DIR1 DIR2
cat >maketest <<EOF
1999-11-01 23:21:16 +08:00
.PATH: $srcdir/config $srcdir/src $srcdir/bin
.c.o:
cp $< H5.o
foo: H5.o
2007-02-07 02:50:04 +08:00
rm -f H5.o
1999-11-01 23:21:16 +08:00
@echo works
EOF
2001-09-28 07:05:16 +08:00
if (MAKE= ${MAKE-make} -f maketest foo) >/dev/null 2>&1; then
SEARCH_RULE='.PATH: '
SEARCH_SEP=' '
AC_MSG_RESULT([.PATH: DIR1 DIR2 ...])
break
fi
dnl No way for make to search directories
SEARCH_RULE='## SEARCH DISABLED: '
SEARCH_SEP=' '
AC_MSG_RESULT([it doesn't])
if test ! -f configure; then
2002-07-18 00:10:46 +08:00
AC_MSG_ERROR([${MAKE-make} requires the build and source directories to be the same])
2001-09-28 07:05:16 +08:00
fi
break
1999-11-01 23:21:16 +08:00
done
rm maketest
2005-02-12 01:40:52 +08:00
dnl ----------------------------------------------------------------------
dnl pmake will throw an error if variables are undefined in a Makefile.
dnl These errors can be changed to warnings using the -V flag.
dnl
AC_SUBST(AM_MAKEFLAGS) AM_MAKEFLAGS=""
2005-02-22 05:30:49 +08:00
dnl Don't run test if MAKE is defined but is the empty string
2005-02-23 01:42:48 +08:00
if test -n "${MAKE-make}"; then
2005-02-22 05:30:49 +08:00
AC_MSG_CHECKING([whether make will build with undefined variables])
2005-02-12 01:40:52 +08:00
cat >maketest <<EOF
foo: \$(UNDEFINED) \$(UNDEFINED2)
@echo \$(UNDEFINED3) works
EOF
if (${MAKE-make} -f maketest foo) >/dev/null 2>&1; then
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no, setting -V flag])
AM_MAKEFLAGS="\-V"
fi
2005-02-22 05:30:49 +08:00
rm maketest
fi
2005-02-12 01:40:52 +08:00
1998-04-03 11:29:38 +08:00
dnl ----------------------------------------------------------------------
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
dnl Production flags? Save the value in $CONFIG_MODE so we have it for
dnl the record.
1998-04-03 11:29:38 +08:00
dnl
AC_MSG_CHECKING(for production mode)
AC_ARG_ENABLE(production,
2002-07-18 00:10:46 +08:00
[AC_HELP_STRING([--enable-production],
[Determines how to run the compiler.])])
1998-04-03 11:29:38 +08:00
1999-08-21 01:05:14 +08:00
case "X-$enable_production" in
2008-05-01 03:51:13 +08:00
X-yes)
2002-02-14 03:57:10 +08:00
enable_production="yes"
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([production])
2001-09-28 07:05:16 +08:00
CONFIG_MODE=production
2006-04-26 05:19:14 +08:00
H5_CFLAGS="$H5_CFLAGS $PROD_CFLAGS"
H5_CPPFLAGS="$H5_CPPFLAGS $PROD_CPPFLAGS"
H5_CXXFLAGS="$H5_CXXFLAGS $PROD_CXXFLAGS"
H5_FCFLAGS="$H5_FCFLAGS $PROD_FCFLAGS"
2001-09-28 07:05:16 +08:00
;;
2008-05-01 03:51:13 +08:00
X-|X-no)
2002-02-14 03:57:10 +08:00
enable_production="no"
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([development])
2001-09-28 07:05:16 +08:00
CONFIG_MODE=development
2006-04-26 05:19:14 +08:00
H5_CFLAGS="$H5_CFLAGS $DEBUG_CFLAGS"
H5_CPPFLAGS="$H5_CPPFLAGS $DEBUG_CPPFLAGS"
H5_CXXFLAGS="$H5_CXXFLAGS $DEBUG_CXXFLAGS"
H5_FCFLAGS="$H5_FCFLAGS $DEBUG_FCFLAGS"
2001-09-28 07:05:16 +08:00
;;
X-pg|X-profile)
2002-02-14 03:57:10 +08:00
enable_production="profile"
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([profile])
2001-09-28 07:05:16 +08:00
CONFIG_MODE=profile
2006-04-26 05:19:14 +08:00
H5_CFLAGS="$H5_CFLAGS $PROFILE_CFLAGS"
H5_CPPFLAGS="$H5_CPPFLAGS $PROFILE_CPPFLAGS"
H5_CXXFLAGS="$H5_CXXFLAGS $PROFILE_CXXFLAGS"
H5_FCFLAGS="$H5_FCFLAGS $PROFILE_FCFLAGS"
2001-09-28 07:05:16 +08:00
;;
*)
2002-02-14 03:57:10 +08:00
enable_production="user-defined"
2002-07-18 00:10:46 +08:00
AC_MSG_RESULT([user-defined])
2001-09-28 07:05:16 +08:00
CONFIG_MODE="$X-enableval"
;;
1998-04-03 11:29:38 +08:00
esac
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
1998-11-13 23:06:06 +08:00
dnl Check for system libraries.
1997-08-16 00:35:53 +08:00
dnl
2002-07-18 00:10:46 +08:00
AC_CHECK_LIB([m], [ceil])
2001-07-06 04:56:01 +08:00
2001-07-18 05:00:32 +08:00
if test "`uname`" = "SunOS" -o "`uname -sr`" = "HP-UX B.11.00"; then
2002-03-21 02:14:36 +08:00
dnl ...for Solaris
2006-10-04 06:41:34 +08:00
AC_CHECK_LIB([socket], [socket])
2002-07-18 00:10:46 +08:00
AC_CHECK_LIB([nsl], [xdr_int])
2001-07-06 04:56:01 +08:00
fi
2002-07-18 00:10:46 +08:00
dnl AC_CHECK_LIB([coug], [main]) dnl ...for ASCI/Red
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
1998-11-13 23:06:06 +08:00
dnl Check for system header files.
1997-08-16 00:35:53 +08:00
dnl
AC_HEADER_STDC
1998-11-03 01:58:28 +08:00
AC_HEADER_TIME
2002-07-18 00:10:46 +08:00
[svn-r6252] Purpose:
Lots of performance improvements & a couple new internal API interfaces.
Description:
Performance Improvements:
- Cached file offset & length sizes in shared file struct, to avoid
constantly looking them up in the FCPL.
- Generic property improvements:
- Added "revision" number to generic property classes to speed
up comparisons.
- Changed method of storing properties from using a hash-table
to the TBBT routines in the library.
- Share the propery names between classes and the lists derived
from them.
- Removed redundant 'def_value' buffer from each property.
- Switching code to use a "copy on write" strategy for
properties in each list, where the properties in each list
are shared with the properties in the class, until a
property's value is changed in a list.
- Fixed error in layout code which was allocating too many buffers.
- Redefined public macros of the form (H5open()/H5check, <variable>)
internally to only be (<variable>), avoiding innumerable useless
calls to H5open() and H5check_version().
- Reuse already zeroed buffers in H5F_contig_fill instead of
constantly re-zeroing them.
- Don't write fill values if writing entire dataset.
- Use gettimeofday() system call instead of time() system when
checking the modification time of a dataset.
- Added reference counted string API and use it for tracking the
names of objects opening in a file (for the ID->name code).
- Removed redundant H5P_get() calls in B-tree routines.
- Redefine H5T datatype macros internally to the library, to avoid
calling H5check redundantly.
- Keep dataspace information for dataset locally instead of reading
from disk each time. Added new module to track open objects
in a file, to allow this (which will be useful eventually for
some FPH5 metadata caching issues).
- Remove H5AC_find macro which was inlining metadata cache lookups,
and call function instead.
- Remove redundant memset() calls from H5G_namei() routine.
- Remove redundant checking of object type when locating objects
in metadata cache and rely on the address only.
- Create default dataset object to use when default dataset creation
property list is used to create datasets, bypassing querying
for all the property list values.
- Use default I/O vector size when performing raw data with the
default dataset transfer property list, instead of querying for
I/O vector size.
- Remove H5P_DEFAULT internally to the library, replacing it with
more specific default property list based on the type of
property list needed.
- Remove redundant memset() calls in object header message (H5O*)
routines.
- Remove redunant memset() calls in data I/O routines.
- Split free-list allocation routines into malloc() and calloc()-
like routines, instead of one combined routine.
- Remove lots of indirection in H5O*() routines.
- Simplify metadata cache entry comparison routine (used when
flushing entire cache out).
- Only enable metadata cache statistics when H5AC_DEBUG is turned
on, instead of always tracking them.
- Simplify address comparison macro (H5F_addr_eq).
- Remove redundant metadata cache entry protections during dataset
creation by protecting the object header once and making all
the modifications necessary for the dataset creation before
unprotecting it.
- Reduce # of "number of element in extent" computations performed
by computing and storing the value during dataspace creation.
- Simplify checking for group location's file information, when file
has not been involving in file-mounting operations.
- Use binary encoding for modification time, instead of ASCII.
- Hoist H5HL_peek calls (to get information in a local heap)
out of loops in many group routine.
- Use static variable for iterators of selections, instead of
dynamically allocation them each time.
- Lookup & insert new entries in one step, avoiding traversing
group's B-tree twice.
- Fixed memory leak in H5Gget_objname_idx() routine (tangential to
performance improvements, but fixed along the way).
- Use free-list for reference counted strings.
- Don't bother copying object names into cached group entries,
since they are re-created when an object is opened.
The benchmark I used to measure these results created several thousand
small (2K) datasets in a file and wrote out the data for them. This is
Elena's "regular.c" benchmark.
These changes resulted in approximately ~4.3x speedup of the
development branch when compared to the previous code in the
development branch and ~1.4x speedup compared to the release
branch.
Additionally, these changes reduce the total memory used (code and
data) by the development branch by ~800KB, bringing the development
branch back into the same ballpark as the release branch.
I'll send out a more detailed description of the benchmark results
as a followup note.
New internal API routines:
Added "reference counted strings" API for tracking strings that get
used by multiple owners without duplicating the strings.
Added "ternary search tree" API for text->object mappings.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
Other platforms/configurations tested?
FreeBSD 4.7 (sleipnir) serial & parallel
Solaris 2.6 (baldric) serial
2003-01-10 01:20:03 +08:00
dnl ----------------------------------------------------------------------
dnl Check for these two functions before <sys/time.h> is checked for, otherwise
dnl they are not detected correctly on Solaris [2.6].
dnl
2004-07-24 01:29:16 +08:00
AC_CHECK_FUNCS(difftime)
AC_CHECK_FUNCS(gettimeofday, [have_gettime="yes"], [have_gettime="no"])
[svn-r6252] Purpose:
Lots of performance improvements & a couple new internal API interfaces.
Description:
Performance Improvements:
- Cached file offset & length sizes in shared file struct, to avoid
constantly looking them up in the FCPL.
- Generic property improvements:
- Added "revision" number to generic property classes to speed
up comparisons.
- Changed method of storing properties from using a hash-table
to the TBBT routines in the library.
- Share the propery names between classes and the lists derived
from them.
- Removed redundant 'def_value' buffer from each property.
- Switching code to use a "copy on write" strategy for
properties in each list, where the properties in each list
are shared with the properties in the class, until a
property's value is changed in a list.
- Fixed error in layout code which was allocating too many buffers.
- Redefined public macros of the form (H5open()/H5check, <variable>)
internally to only be (<variable>), avoiding innumerable useless
calls to H5open() and H5check_version().
- Reuse already zeroed buffers in H5F_contig_fill instead of
constantly re-zeroing them.
- Don't write fill values if writing entire dataset.
- Use gettimeofday() system call instead of time() system when
checking the modification time of a dataset.
- Added reference counted string API and use it for tracking the
names of objects opening in a file (for the ID->name code).
- Removed redundant H5P_get() calls in B-tree routines.
- Redefine H5T datatype macros internally to the library, to avoid
calling H5check redundantly.
- Keep dataspace information for dataset locally instead of reading
from disk each time. Added new module to track open objects
in a file, to allow this (which will be useful eventually for
some FPH5 metadata caching issues).
- Remove H5AC_find macro which was inlining metadata cache lookups,
and call function instead.
- Remove redundant memset() calls from H5G_namei() routine.
- Remove redundant checking of object type when locating objects
in metadata cache and rely on the address only.
- Create default dataset object to use when default dataset creation
property list is used to create datasets, bypassing querying
for all the property list values.
- Use default I/O vector size when performing raw data with the
default dataset transfer property list, instead of querying for
I/O vector size.
- Remove H5P_DEFAULT internally to the library, replacing it with
more specific default property list based on the type of
property list needed.
- Remove redundant memset() calls in object header message (H5O*)
routines.
- Remove redunant memset() calls in data I/O routines.
- Split free-list allocation routines into malloc() and calloc()-
like routines, instead of one combined routine.
- Remove lots of indirection in H5O*() routines.
- Simplify metadata cache entry comparison routine (used when
flushing entire cache out).
- Only enable metadata cache statistics when H5AC_DEBUG is turned
on, instead of always tracking them.
- Simplify address comparison macro (H5F_addr_eq).
- Remove redundant metadata cache entry protections during dataset
creation by protecting the object header once and making all
the modifications necessary for the dataset creation before
unprotecting it.
- Reduce # of "number of element in extent" computations performed
by computing and storing the value during dataspace creation.
- Simplify checking for group location's file information, when file
has not been involving in file-mounting operations.
- Use binary encoding for modification time, instead of ASCII.
- Hoist H5HL_peek calls (to get information in a local heap)
out of loops in many group routine.
- Use static variable for iterators of selections, instead of
dynamically allocation them each time.
- Lookup & insert new entries in one step, avoiding traversing
group's B-tree twice.
- Fixed memory leak in H5Gget_objname_idx() routine (tangential to
performance improvements, but fixed along the way).
- Use free-list for reference counted strings.
- Don't bother copying object names into cached group entries,
since they are re-created when an object is opened.
The benchmark I used to measure these results created several thousand
small (2K) datasets in a file and wrote out the data for them. This is
Elena's "regular.c" benchmark.
These changes resulted in approximately ~4.3x speedup of the
development branch when compared to the previous code in the
development branch and ~1.4x speedup compared to the release
branch.
Additionally, these changes reduce the total memory used (code and
data) by the development branch by ~800KB, bringing the development
branch back into the same ballpark as the release branch.
I'll send out a more detailed description of the benchmark results
as a followup note.
New internal API routines:
Added "reference counted strings" API for tracking strings that get
used by multiple owners without duplicating the strings.
Added "ternary search tree" API for text->object mappings.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
Other platforms/configurations tested?
FreeBSD 4.7 (sleipnir) serial & parallel
Solaris 2.6 (baldric) serial
2003-01-10 01:20:03 +08:00
1998-12-15 00:19:21 +08:00
dnl Unix
2002-07-18 00:10:46 +08:00
AC_CHECK_HEADERS([sys/resource.h sys/time.h unistd.h sys/ioctl.h sys/stat.h])
AC_CHECK_HEADERS([sys/socket.h sys/types.h])
AC_CHECK_HEADERS([stddef.h setjmp.h features.h])
AC_CHECK_HEADERS([stdint.h], [C9x=yes])
1998-12-15 00:19:21 +08:00
dnl Windows
2005-09-02 04:38:18 +08:00
case "`uname`" in
CYGWIN*)
AC_CHECK_HEADERS([io.h sys/timeb.h])
;;
*)
AC_CHECK_HEADERS([io.h winsock.h sys/timeb.h])
;;
esac
2000-09-15 06:19:24 +08:00
case "$host" in
2001-09-28 07:05:16 +08:00
alpha*-dec*-osf*)
dnl The <sys/sysinfo.h> and <sys/proc.h> are needed on the DEC
dnl Alpha to turn off UAC fixing. We do *not* attempt to
dnl locate these files on other systems because there are too
dnl many problems with including them.
2002-07-18 00:10:46 +08:00
AC_CHECK_HEADERS([sys/sysinfo.h sys/proc.h])
2001-09-28 07:05:16 +08:00
;;
2003-11-25 00:47:18 +08:00
mips*-sgi*-irix*)
dnl The <sys/fpu.h> is needed on the SGI machines to turn off
dnl denormalized floating-point values going to zero. We do *not*
dnl attempt to dnl locate these files on other systems because there
dnl may be problems with including them.
AC_CHECK_HEADERS([sys/fpu.h])
AC_CHECK_FUNCS([get_fpc_csr])
;;
1999-03-03 01:15:35 +08:00
esac
1997-08-16 00:35:53 +08:00
2011-04-15 05:21:59 +08:00
dnl ----------------------------------------------------------------------
dnl Some platforms require that all symbols are resolved when a library
dnl is linked. We can use the -no-undefined flag to tell libtool that
dnl it will be able to build shared libraries on these architectures,
dnl as it will not do so by default.
dnl
if test "X${enable_shared}" = "Xyes"; then
AC_MSG_CHECKING([if libtool needs -no-undefined flag to build shared libraries])
case "`uname`" in
CYGWIN*|MINGW*|AIX*)
dnl Add in the -no-undefined flag to LDFLAGS for libtool.
AC_MSG_RESULT([yes])
H5_LDFLAGS="$H5_LDFLAGS -no-undefined"
;;
*)
dnl Don't add in anything.
AC_MSG_RESULT([no])
;;
esac
fi
2002-07-18 00:10:46 +08:00
dnl ----------------------------------------------------------------------
2010-05-06 01:34:26 +08:00
dnl Test for Largefile support.
dnl
AC_MSG_CHECKING([if configure should try to set up large file support])
AC_ARG_ENABLE([largefile],
[AC_HELP_STRING([--disable-largefile],
[omit support for large files])])
dnl If largefile support is enabled, then set up appropriate compiler options.
if test "$enable_largefile" != no; then
AC_MSG_RESULT(yes)
dnl Check for needed compiler options. This check is pulled drectly
dnl from autoconf's AC_SYS_LARGEFILE macro, as of Autoconf v2.65.
AC_CACHE_CHECK([for special C compiler options needed for large files],
ac_cv_sys_largefile_CC,
[ac_cv_sys_largefile_CC=no
if test "$GCC" != yes; then
ac_save_CC=$CC
while :; do
# IRIX 6.2 and later do not support large files by default,
# so use the C compiler's -n32 option if that helps.
AC_LANG_CONFTEST([AC_LANG_PROGRAM([_AC_SYS_LARGEFILE_TEST_INCLUDES])])
AC_COMPILE_IFELSE([], [break])
CC="$CC -n32"
AC_COMPILE_IFELSE([], [ac_cv_sys_largefile_CC=' -n32'; break])
break
done
CC=$ac_save_CC
rm -f conftest.$ac_ext
fi])
if test "$ac_cv_sys_largefile_CC" != no; then
CC=$CC$ac_cv_sys_largefile_CC
fi
dnl Use the macro _AC_SYS_LARGEFILE_MACRO_VALUE to test defines
dnl that might need to be set for largefile support to behave
dnl correctly. This macro is defined in acsite.m4 and overrides
dnl the version provided by Autoconf (as of v2.65). The custom
dnl macro additionally adds the appropriate defines to AM_CPPFLAGS
dnl so that later configure checks have them visible.
dnl Check for _FILE_OFFSET_BITS
_AC_SYS_LARGEFILE_MACRO_VALUE(_FILE_OFFSET_BITS, 64,
ac_cv_sys_file_offset_bits,
[Number of bits in a file offset, on hosts where this is settable.],
[_AC_SYS_LARGEFILE_TEST_INCLUDES])
dnl Check for _LARGE_FILES
if test $ac_cv_sys_file_offset_bits = unknown; then
_AC_SYS_LARGEFILE_MACRO_VALUE(_LARGE_FILES, 1,
ac_cv_sys_large_files,
[Define for large files, on AIX-style hosts.],
[_AC_SYS_LARGEFILE_TEST_INCLUDES])
fi
dnl Now actually test to see if we can create large files after we've
dnl checked for any needed defines.
AC_MSG_CHECKING([if large (64-bit) files are supported on this system.])
AC_CACHE_VAL([hdf5_cv_have_lfs],
[AC_TRY_RUN([
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <fcntl.h>
#define BIG_FILE (off_t)0x80000000UL
int main(void) {
int fd;
if ((fd=open("test.conf", O_RDWR|O_TRUNC|O_CREAT, 0666)) < 0) exit(1);
if (lseek(fd, BIG_FILE, SEEK_SET)!=BIG_FILE) exit(1);
if (5!=write(fd, "hello", (size_t)5)) exit(1);
if (lseek(fd, 2*BIG_FILE, SEEK_SET) != 2*BIG_FILE) exit(1);
if (5!=write(fd, "hello", (size_t)5)) exit(1);
if (unlink("test.conf") < 0) exit(1);
exit(0);
}
],[hdf5_cv_have_lfs=yes],[hdf5_cv_have_lfs=no],)])
if test "X${hdf5_cv_have_lfs}" = "Xyes"; then
AC_MSG_RESULT([yes])
LARGEFILE="yes"
fi
if test "X${hdf5_cv_have_lfs}" = "Xno"; then
AC_MSG_RESULT([no])
LARGEFILE="no"
fi
else
LARGEFILE="no"
AC_MSG_RESULT(no)
fi
dnl ----------------------------------------------------------------------
dnl Add necessary defines for Linux Systems.
2002-07-18 00:10:46 +08:00
dnl
2001-05-09 04:59:31 +08:00
case "$host_cpu-$host_vendor-$host_os" in
2001-09-28 07:05:16 +08:00
*linux*)
2010-05-06 01:34:26 +08:00
dnl If largefile support is enabled, then make available various
dnl LFS-related routines using the following _LARGEFILE*_SOURCE macros.
if test "X$LARGEFILE" != "Xno"; then
AM_CPPFLAGS="-D_LARGEFILE64_SOURCE -D_LARGEFILE_SOURCE $AM_CPPFLAGS"
2001-09-28 07:05:16 +08:00
fi
dnl Add POSIX support on Linux systems, so <features.h> defines
dnl __USE_POSIX, which is required to get the prototype for fdopen
2010-06-05 00:01:39 +08:00
dnl defined correctly in <stdio.h>.
dnl This flag was removed from h5cc as of 2009-10-17 when it was found
dnl that the flag broke compiling netCDF-4 code with h5cc, but kept in
dnl H5_CPPFLAGS because fdopen and HDfdopen fail without it. HDfdopen
dnl is used only by H5_debug_mask which is used only when debugging in
dnl H5_init_library (all in H5.c). When the flag was removed this was
dnl the only compile failure noted; however, defining the Macro
dnl makes available the functionality from the POSIX.1 standard as well
dnl as all or the ISO C facilities, according to a discussion of
dnl defining the _POSIX_SOURCE, _BSD_SOURCE, and _POSIX_C_SOURCE Macros
dnl and their interaction found at
dnl http://www.gnu.org/s/libc/manual/html_node/Feature-Test-Macros.html.
2009-10-23 02:38:21 +08:00
H5_CPPFLAGS="-D_POSIX_SOURCE $H5_CPPFLAGS"
2001-07-17 00:40:56 +08:00
2001-09-28 07:05:16 +08:00
dnl Also add BSD support on Linux systems, so <features.h> defines
dnl __USE_BSD, which is required to get the prototype for strdup
dnl defined correctly in <string.h> and snprintf & vsnprintf defined
dnl correctly in <stdio.h>
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CPPFLAGS="-D_BSD_SOURCE $AM_CPPFLAGS"
2001-09-28 07:05:16 +08:00
;;
2001-05-09 04:59:31 +08:00
esac
2011-05-24 01:02:39 +08:00
dnl Need to add the AM_ and H5_ into CPFLAGS/CPPFLAGS to make them visible
dnl for configure checks.
dnl Note: Both will be restored by the end of configure.
CPPFLAGS="$H5_CPPFLAGS $AM_CPPFLAGS $CPPFLAGS"
CFLAGS="$H5_CFLAGS $AM_CFLAGS $CFLAGS"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2001-05-09 04:59:31 +08:00
AC_TRY_COMPILE([#include <sys/types.h>],
2002-07-18 00:10:46 +08:00
[off64_t n = 0;],
2010-08-10 06:13:20 +08:00
[AC_CHECK_FUNCS([lseek64 fseeko64 ftello64 ftruncate64])],
[AC_MSG_RESULT([skipping test for lseek64(), fseeko64 , ftello64, ftruncate64() because off64_t is not defined])])
2010-08-07 03:32:23 +08:00
AC_CHECK_FUNCS(fseeko ftello)
2006-05-14 13:09:25 +08:00
AC_TRY_COMPILE([
#include <sys/types.h>
#include <sys/stat.h>],
[struct stat64 sb;],
[AC_CHECK_FUNCS([stat64 fstat64])],
[AC_MSG_RESULT([skipping test for stat64() and fstat64()])])
2001-05-09 04:59:31 +08:00
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
[svn-r339] Changes since 19980408
----------------------
./src/H5Osdspace.c
./html/H5.format.html
In the past we were allowed to have >2GB files on a 32-bit
machine as long as no dataset within the file was larger than
4GB (or whatever sizeof(size_t) is). That's been fixed now.
All dataset size calculations are done with `hsize_t' which is
normally defined as `unsigned long long'.
./src/H5F.c
./src/H5Ffamily.c
./src/H5Fprivate.h
./src/H5P.c
./src/H5Ppublic.h
The file family member size can now be set/queried. The
default is still 64MB, but it can be set to 1GB by saying:
H5Pset_family (plist, 30, H5P_DEFAULT);
When opening an existing file family the specified
bits-per-member is ignored and the first member of the family
determines the bits-per-member, which can be retrieved with
H5Pget_family().
./acconfig.h
./configure.in
./src/H5config.h
./src/H5public.h
Added `--disable-hsizet' so that those with old GCC compilers
(<2.8.1) can still compile the code.
./src/H5.c
./src/H5private.h
Added HDfprintf() which works just like fprintf() except you
can give `H' as a size modifier for the integer conversions
and supply an `hsize_t' or `hssize_t' argument without casting
it. For instance:
hsize_t npoints = H5Sget_npoints(space);
HDfprintf(stdout,"Dataset has %Hd (%#018Hx) points\n",
npoints, npoints);
You can now give `%a' as a format to print an address, but all
formating flags are ignored and it causes the return value of
HDfprintf() to not include the characters in the address (but
who uses the return value anyway :-). Example:
H5G_t *grp;
HDfprintf(stdout, "Group object header at %a\n",
&(grp->ent.header));
Added HDstrtoll() which works exactly like [HD]strtol() except
the result is an int64.
./src/debug.c
Large addresses can now be entered from the command-line. Use
either decimal, octal (leading `0') or hexadecimal (leading
`0x') when giving the address.
./src/h5ls.c
The printf format for dataset dimensions was changed to `%Hu'
to support large datasets.
./test/big.c [NEW]
A test for big datasets on 32-bit machines. This test is not
run by default. Don't try to run it on an nfs-mounted file
system or other file system that doesn't support holes because
it creates two 32GB datasets of all zero.
1998-04-10 04:22:11 +08:00
dnl Data types and their sizes.
1997-08-16 00:35:53 +08:00
dnl
AC_TYPE_OFF_T
2002-07-18 00:10:46 +08:00
AC_CHECK_TYPE([size_t], [unsigned long])
AC_CHECK_TYPE([ssize_t], [long])
[svn-r18634] Description:
Clean up compiler warnings.
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (amani) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.6.3 (amazon) in debug mode
Mac OS X/32 10.6.3 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2010-04-28 02:25:31 +08:00
AC_CHECK_TYPE([ptrdiff_t], [long])
1997-08-16 00:35:53 +08:00
AC_C_BIGENDIAN
2002-07-18 00:10:46 +08:00
AC_CHECK_SIZEOF([char], [1])
AC_CHECK_SIZEOF([short], [2])
AC_CHECK_SIZEOF([int], [4])
2009-04-24 02:25:16 +08:00
AC_CHECK_SIZEOF([unsigned], [4])
2002-07-18 00:10:46 +08:00
AC_CHECK_SIZEOF([long], [4])
AC_CHECK_SIZEOF([long long], [8])
AC_CHECK_SIZEOF([__int64], [8])
AC_CHECK_SIZEOF([float], [4])
AC_CHECK_SIZEOF([double], [8])
AC_CHECK_SIZEOF([long double], [8])
1998-11-19 02:40:09 +08:00
2000-04-21 02:41:51 +08:00
dnl Checkpoint the cache
AC_CACHE_SAVE
1999-06-11 23:53:15 +08:00
dnl Posix.1g types (C9x)
1998-11-19 02:40:09 +08:00
cat >>confdefs.h <<\EOF
#include <sys/types.h>
EOF
2000-09-15 06:19:24 +08:00
if test "X$C9x" = "Xyes"; then
2001-09-28 07:05:16 +08:00
cat >>confdefs.h <<\EOF
1999-08-21 01:05:14 +08:00
#include <stdint.h>
EOF
fi
2002-07-18 00:10:46 +08:00
AC_CHECK_SIZEOF( int8_t, [1])
AC_CHECK_SIZEOF( uint8_t, [1])
AC_CHECK_SIZEOF( int_least8_t, [1])
AC_CHECK_SIZEOF( uint_least8_t, [1])
AC_CHECK_SIZEOF( int_fast8_t, [1])
AC_CHECK_SIZEOF( uint_fast8_t, [1])
AC_CHECK_SIZEOF( int16_t, [2])
AC_CHECK_SIZEOF( uint16_t, [2])
AC_CHECK_SIZEOF( int_least16_t, [2])
AC_CHECK_SIZEOF(uint_least16_t, [2])
AC_CHECK_SIZEOF( int_fast16_t, [2])
AC_CHECK_SIZEOF( uint_fast16_t, [2])
AC_CHECK_SIZEOF( int32_t, [4])
AC_CHECK_SIZEOF( uint32_t, [4])
AC_CHECK_SIZEOF( int_least32_t, [4])
AC_CHECK_SIZEOF(uint_least32_t, [4])
AC_CHECK_SIZEOF( int_fast32_t, [4])
AC_CHECK_SIZEOF( uint_fast32_t, [4])
AC_CHECK_SIZEOF( int64_t, [8])
AC_CHECK_SIZEOF( uint64_t, [8])
AC_CHECK_SIZEOF( int_least64_t, [8])
AC_CHECK_SIZEOF(uint_least64_t, [8])
AC_CHECK_SIZEOF( int_fast64_t, [8])
AC_CHECK_SIZEOF( uint_fast64_t, [8])
AC_CHECK_SIZEOF([size_t], [4])
AC_CHECK_SIZEOF([ssize_t], [4])
[svn-r18634] Description:
Clean up compiler warnings.
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (amani) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.6.3 (amazon) in debug mode
Mac OS X/32 10.6.3 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2010-04-28 02:25:31 +08:00
AC_CHECK_SIZEOF([ptrdiff_t], [4])
2002-07-18 00:10:46 +08:00
1998-03-06 05:27:38 +08:00
cat >>confdefs.h <<\EOF
#include <sys/types.h> /*for off_t definition*/
EOF
2002-07-18 00:10:46 +08:00
AC_CHECK_SIZEOF([off_t], [4])
2006-05-11 06:29:25 +08:00
AC_CHECK_SIZEOF([off64_t], [8])
1997-08-16 00:35:53 +08:00
2000-04-21 02:41:51 +08:00
dnl Checkpoint the cache
AC_CACHE_SAVE
2001-07-18 05:24:38 +08:00
dnl ----------------------------------------------------------------------
dnl Check if the dev_t type is a scalar type (must come after the check for
dnl sys/types.h)
2001-09-28 07:05:16 +08:00
AC_MSG_CHECKING([if dev_t is scalar])
AC_TRY_COMPILE([
2001-07-18 05:24:38 +08:00
#ifdef HAVE_SYS_TYPES_H
#include <sys/types.h>
#endif
2002-04-23 07:44:41 +08:00
],
[dev_t d1, d2; if(d1==d2) return 0;],
2002-07-18 00:10:46 +08:00
AC_DEFINE([DEV_T_IS_SCALAR], [1],
2005-02-01 11:17:02 +08:00
[Define if `dev_t' is a scalar])
2002-04-23 07:44:41 +08:00
AC_MSG_RESULT(yes),
AC_MSG_RESULT(no)
2001-09-28 07:05:16 +08:00
)
2001-07-18 05:24:38 +08:00
2002-02-14 03:57:10 +08:00
dnl ----------------------------------------------------------------------
dnl Fake --with-xxx option to allow us to create a help message for the
dnl following --with-xxx options which can take either a =DIR or =INC,LIB
dnl specifier.
dnl
2002-07-18 00:10:46 +08:00
AC_ARG_WITH([fnord],
2002-02-14 03:57:10 +08:00
[
For the following --with-xxx options, you can specify where the header
files and libraries are in two different ways:
--with-xxx=INC,LIB - Specify individually the include directory and
library directory separated by a comma
--with-xxx=DIR - Specify only the directory which contains the
include/ and lib/ subdirectories
2002-07-18 00:10:46 +08:00
])
2002-02-14 03:57:10 +08:00
2002-07-13 02:29:41 +08:00
dnl ----------------------------------------------------------------------
dnl Is the dmalloc present? It has a header file `dmalloc.h' and a library
dnl `-ldmalloc' and their locations might be specified with the `--with-dmalloc'
dnl command-line switch. The value is an include path and/or a library path.
dnl If the library path is specified then it must be preceded by a comma.
dnl
2002-07-18 00:10:46 +08:00
AC_ARG_WITH([dmalloc],
[AC_HELP_STRING([--with-dmalloc=DIR],
[Use dmalloc memory debugging aid [default=no]])],,
withval=no)
2002-07-13 02:29:41 +08:00
case $withval in
yes)
HAVE_DMALLOC="yes"
AC_CHECK_HEADERS(dmalloc.h)
AC_CHECK_LIB(dmalloc, dmalloc_shutdown,, unset HAVE_DMALLOC)
if test -z "$HAVE_DMALLOC" -a -n "$HDF5_CONFIG_ABORT"; then
AC_MSG_ERROR(couldn't find dmalloc library)
fi
;;
no)
HAVE_DMALLOC="no"
AC_MSG_CHECKING(for dmalloc library)
AC_MSG_RESULT(suppressed)
;;
*)
HAVE_DMALLOC="yes"
case "$withval" in
*,*)
dmalloc_inc="`echo $withval |cut -f1 -d,`"
dmalloc_lib="`echo $withval |cut -f2 -d, -s`"
;;
*)
if test -n "$withval"; then
dmalloc_inc="$withval/include"
dmalloc_lib="$withval/lib"
fi
;;
esac
2003-01-24 03:19:02 +08:00
dnl Trying to include -I/usr/include and -L/usr/lib is redundant and
dnl can mess some compilers up.
if test "X$dmalloc_inc" = "X/usr/include"; then
dmalloc_inc=""
fi
if test "X$dmalloc_lib" = "X/usr/lib"; then
dmalloc_lib=""
fi
2002-07-13 02:29:41 +08:00
saved_CPPFLAGS="$CPPFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_CPPFLAGS="$AM_CPPFLAGS"
2002-07-13 02:29:41 +08:00
saved_LDFLAGS="$LDFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_LDFLAGS="$AM_LDFLAGS"
2002-07-13 02:29:41 +08:00
if test -n "$dmalloc_inc"; then
CPPFLAGS="$CPPFLAGS -I$dmalloc_inc"
2009-10-15 05:43:05 +08:00
AM_CPPFLAGS="$AM_CPPFLAGS -I$dmalloc_inc"
2002-07-13 02:29:41 +08:00
fi
2009-10-15 05:43:05 +08:00
AC_CHECK_HEADERS(dmalloc.h,,CPPFLAGS="$saved_CPPFLAGS"; AM_CPPFLAGS="$saved_AM_CPPFLAGS")
2002-07-13 02:29:41 +08:00
if test -n "$dmalloc_lib"; then
LDFLAGS="$LDFLAGS -L$dmalloc_lib"
2009-10-15 05:43:05 +08:00
AM_LDFLAGS="$AM_LDFLAGS -L$dmalloc_lib"
2002-07-13 02:29:41 +08:00
fi
2009-10-15 05:43:05 +08:00
AC_CHECK_LIB(dmalloc, dmalloc_shutdown,, LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset HAVE_DMALLOC)
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2002-07-13 02:29:41 +08:00
if test -z "$HAVE_DMALLOC" -a -n "$HDF5_CONFIG_ABORT"; then
AC_MSG_ERROR(couldn't find dmalloc library)
fi
;;
esac
[svn-r1525] configure.in:
Added the checking for zlib and jpeg libraries when doing the -with-hdf4
option. HDF4 applications in general needs lib mfhdf, df, z, and jpeg.
Removed the earlier alone test for jpeg. It conflicted with the hdf4
tests. Also, moved the hdf5-own library test for zlib after the hdf4
tests so that when hdf4 is used, the results would cover the zlib
test.
configure:
derived from configure.in by autoconf.
Platforms tested:
IRIX 6.5, IRIX64 6.5, Solaris 2.6 (with default gcc, tested h5toh4
tool).
Comments:
This still generates compile commands that listed the "-lmfhdf ..."
library list before the -L/usr/ncsa/lib. I traced the problem to
the libtool. (I stopped at the door and dared not to enter this
libtool cave. :-) Most compilers, except solaris' cc, tolerate
the "duplicated" library list.
1999-07-22 09:54:16 +08:00
dnl ----------------------------------------------------------------------
2001-07-31 06:48:40 +08:00
dnl Is the GNU zlib present? It has a header file `zlib.h' and a library
2001-01-04 14:34:06 +08:00
dnl `-lz' and their locations might be specified with the `--with-zlib'
2001-07-31 06:48:40 +08:00
dnl command-line switch. The value is an include path and/or a library path.
[svn-r1525] configure.in:
Added the checking for zlib and jpeg libraries when doing the -with-hdf4
option. HDF4 applications in general needs lib mfhdf, df, z, and jpeg.
Removed the earlier alone test for jpeg. It conflicted with the hdf4
tests. Also, moved the hdf5-own library test for zlib after the hdf4
tests so that when hdf4 is used, the results would cover the zlib
test.
configure:
derived from configure.in by autoconf.
Platforms tested:
IRIX 6.5, IRIX64 6.5, Solaris 2.6 (with default gcc, tested h5toh4
tool).
Comments:
This still generates compile commands that listed the "-lmfhdf ..."
library list before the -L/usr/ncsa/lib. I traced the problem to
the libtool. (I stopped at the door and dared not to enter this
libtool cave. :-) Most compilers, except solaris' cc, tolerate
the "duplicated" library list.
1999-07-22 09:54:16 +08:00
dnl If the library path is specified then it must be preceded by a comma.
dnl
2004-06-09 21:47:20 +08:00
AC_SUBST(USE_FILTER_DEFLATE) USE_FILTER_DEFLATE="no"
2002-07-18 00:10:46 +08:00
AC_ARG_WITH([zlib],
[AC_HELP_STRING([--with-zlib=DIR],
2003-04-05 01:54:25 +08:00
[Use zlib library for external deflate I/O
filter [default=yes]])],,
2002-07-18 00:10:46 +08:00
withval=yes)
2001-01-04 14:34:06 +08:00
case $withval in
2001-07-31 06:48:40 +08:00
yes)
2001-09-19 03:26:46 +08:00
HAVE_ZLIB="yes"
2002-07-18 00:10:46 +08:00
AC_CHECK_HEADERS([zlib.h], [HAVE_ZLIB_H="yes"])
AC_CHECK_LIB([z], [compress2],, [unset HAVE_ZLIB])
AC_CHECK_FUNC([compress2], [HAVE_COMPRESS2="yes"])
2002-04-23 07:44:41 +08:00
if test -z "$HAVE_ZLIB" -a -n "$HDF5_CONFIG_ABORT"; then
2002-07-18 00:10:46 +08:00
AC_MSG_ERROR([couldn't find zlib library])
2002-04-23 07:44:41 +08:00
fi
2001-07-31 06:48:40 +08:00
;;
no)
2001-09-19 03:26:46 +08:00
HAVE_ZLIB="no"
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for GNU zlib])
AC_MSG_RESULT([suppressed])
2001-07-31 06:48:40 +08:00
;;
*)
2001-09-19 03:26:46 +08:00
HAVE_ZLIB="yes"
2001-08-11 05:34:40 +08:00
case "$withval" in
*,*)
zlib_inc="`echo $withval |cut -f1 -d,`"
zlib_lib="`echo $withval |cut -f2 -d, -s`"
;;
*)
if test -n "$withval"; then
zlib_inc="$withval/include"
zlib_lib="$withval/lib"
fi
;;
esac
2003-01-24 03:19:02 +08:00
dnl Trying to include -I/usr/include and -L/usr/lib is redundant and
dnl can mess some compilers up.
if test "X$zlib_inc" = "X/usr/include"; then
zlib_inc=""
fi
if test "X$zlib_lib" = "X/usr/lib"; then
zlib_lib=""
fi
2001-08-11 05:34:40 +08:00
saved_CPPFLAGS="$CPPFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_CPPFLAGS="$AM_CPPFLAGS"
2001-08-11 05:34:40 +08:00
saved_LDFLAGS="$LDFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_LDFLAGS="$AM_LDFLAGS"
2001-08-11 05:34:40 +08:00
2001-07-31 06:48:40 +08:00
if test -n "$zlib_inc"; then
CPPFLAGS="$CPPFLAGS -I$zlib_inc"
2009-10-15 05:43:05 +08:00
AM_CPPFLAGS="$AM_CPPFLAGS -I$zlib_inc"
2001-07-31 06:48:40 +08:00
fi
2002-07-18 00:10:46 +08:00
AC_CHECK_HEADERS([zlib.h],
2009-10-15 05:43:05 +08:00
[HAVE_ZLIB_H="yes"],
[CPPFLAGS="$saved_CPPFLAGS"; AM_CPPFLAGS="$saved_AM_CPPFLAGS"])
2001-07-31 06:48:40 +08:00
if test -n "$zlib_lib"; then
LDFLAGS="$LDFLAGS -L$zlib_lib"
2009-10-15 05:43:05 +08:00
AM_LDFLAGS="$AM_LDFLAGS -L$zlib_lib"
2001-07-31 06:48:40 +08:00
fi
2001-08-11 05:34:40 +08:00
2002-07-18 00:10:46 +08:00
AC_CHECK_LIB([z], [compress2],,
2009-10-15 05:43:05 +08:00
[LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset HAVE_ZLIB])
2002-07-18 00:10:46 +08:00
AC_CHECK_FUNC([compress2], [HAVE_COMPRESS2="yes"])
2002-04-23 07:44:41 +08:00
if test -z "$HAVE_ZLIB" -a -n "$HDF5_CONFIG_ABORT"; then
2002-07-18 00:10:46 +08:00
AC_MSG_ERROR([couldn't find zlib library])
2002-04-23 07:44:41 +08:00
fi
2001-07-31 06:48:40 +08:00
;;
2001-01-04 14:34:06 +08:00
esac
1998-11-13 23:06:06 +08:00
2002-07-18 00:10:46 +08:00
if test "x$HAVE_ZLIB" = "xyes" -a "x$HAVE_ZLIB_H" = "xyes" -a "x$HAVE_COMPRESS2" = "xyes"; then
2005-05-21 01:32:49 +08:00
AC_DEFINE([HAVE_FILTER_DEFLATE], [1], [Define if support for deflate (zlib) filter is enabled])
2004-06-09 21:47:20 +08:00
USE_FILTER_DEFLATE="yes"
2003-04-05 01:54:25 +08:00
dnl Add "deflate" to external filter list
if test "X$EXTERNAL_FILTERS" != "X"; then
EXTERNAL_FILTERS="${EXTERNAL_FILTERS},"
fi
2005-05-21 01:32:49 +08:00
EXTERNAL_FILTERS="${EXTERNAL_FILTERS}deflate(zlib)"
2003-04-05 01:54:25 +08:00
fi
dnl ----------------------------------------------------------------------
dnl Is the szlib present? It has a header file `szlib.h' and a library
dnl `-lsz' and their locations might be specified with the `--with-szlib'
dnl command-line switch. The value is an include path and/or a library path.
dnl If the library path is specified then it must be preceded by a comma.
dnl
2004-06-09 21:47:20 +08:00
AC_SUBST(USE_FILTER_SZIP) USE_FILTER_SZIP="no"
2003-04-05 01:54:25 +08:00
AC_ARG_WITH([szlib],
[AC_HELP_STRING([--with-szlib=DIR],
[Use szlib library for external szlib I/O
2003-05-16 07:25:38 +08:00
filter [default=no]])],,
withval=no)
2003-04-05 01:54:25 +08:00
case $withval in
yes)
HAVE_SZLIB="yes"
AC_CHECK_HEADERS([szlib.h], [HAVE_SZLIB_H="yes"])
AC_CHECK_LIB([sz], [SZ_BufftoBuffCompress],, [unset HAVE_SZLIB])
if test -z "$HAVE_SZLIB" -a -n "$HDF5_CONFIG_ABORT"; then
AC_MSG_ERROR([couldn't find szlib library])
fi
;;
no)
HAVE_SZLIB="no"
AC_MSG_CHECKING([for szlib])
AC_MSG_RESULT([suppressed])
;;
*)
HAVE_SZLIB="yes"
case "$withval" in
*,*)
szlib_inc="`echo $withval |cut -f1 -d,`"
szlib_lib="`echo $withval |cut -f2 -d, -s`"
;;
*)
if test -n "$withval"; then
szlib_inc="$withval/include"
szlib_lib="$withval/lib"
fi
;;
esac
2007-03-09 00:57:36 +08:00
2003-04-05 01:54:25 +08:00
dnl Trying to include -I/usr/include and -L/usr/lib is redundant and
dnl can mess some compilers up.
if test "X$szlib_inc" = "X/usr/include"; then
szlib_inc=""
2007-03-09 00:57:36 +08:00
fi
2003-04-05 01:54:25 +08:00
if test "X$szlib_lib" = "X/usr/lib"; then
szlib_lib=""
fi
2007-03-09 00:57:36 +08:00
2003-04-05 01:54:25 +08:00
saved_CPPFLAGS="$CPPFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_CPPFLAGS="$AM_CPPFLAGS"
2003-04-05 01:54:25 +08:00
saved_LDFLAGS="$LDFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_LDFLAGS="$AM_LDFLAGS"
2007-03-09 00:57:36 +08:00
2003-04-05 01:54:25 +08:00
if test -n "$szlib_inc"; then
CPPFLAGS="$CPPFLAGS -I$szlib_inc"
2009-10-15 05:43:05 +08:00
AM_CPPFLAGS="$AM_CPPFLAGS -I$szlib_inc"
2003-04-05 01:54:25 +08:00
fi
2007-03-09 00:57:36 +08:00
2003-04-05 01:54:25 +08:00
AC_CHECK_HEADERS([szlib.h],
2009-10-15 05:43:05 +08:00
[HAVE_SZLIB_H="yes"],
[CPPFLAGS="$saved_CPPFLAGS"; AM_CPPFLAGS="$saved_AM_CPPFLAGS"])
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2003-04-05 01:54:25 +08:00
if test -n "$szlib_lib"; then
LDFLAGS="$LDFLAGS -L$szlib_lib"
2009-10-15 05:43:05 +08:00
AM_LDFLAGS="$AM_LDFLAGS -L$szlib_lib"
2003-04-05 01:54:25 +08:00
fi
2007-03-09 00:57:36 +08:00
2003-04-05 01:54:25 +08:00
AC_CHECK_LIB([sz], [SZ_BufftoBuffCompress],,
2009-10-15 05:43:05 +08:00
[LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset HAVE_SZLIB])
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2003-04-05 01:54:25 +08:00
if test -z "$HAVE_SZLIB" -a -n "$HDF5_CONFIG_ABORT"; then
AC_MSG_ERROR([couldn't find szlib library])
fi
;;
esac
if test "x$HAVE_SZLIB" = "xyes" -a "x$HAVE_SZLIB_H" = "xyes"; then
2004-07-02 01:38:04 +08:00
dnl SZLIB library is available. Check if it can encode
AC_MSG_CHECKING([for szlib encoder])
2008-10-25 01:19:34 +08:00
dnl Set LD_LIBRARY_PATH so encoder test can find the library and run.
dnl Also add LL_PATH substitution to Makefiles so they can use the
dnl path as well, for testing examples.
2008-10-25 05:30:07 +08:00
if test -z "$LD_LIBRARY_PATH"; then
2008-10-25 01:19:34 +08:00
export LD_LIBRARY_PATH="$szlib_lib"
2008-10-25 05:30:07 +08:00
else
export LD_LIBRARY_PATH="$szlib_lib:$LD_LIBRARY_PATH"
fi
AC_SUBST([LL_PATH]) LL_PATH="$LD_LIBRARY_PATH"
2007-03-09 00:57:36 +08:00
2004-07-02 01:38:04 +08:00
AC_CACHE_VAL([hdf5_cv_szlib_can_encode],
[AC_TRY_RUN([
#include <szlib.h>
2007-03-09 00:57:36 +08:00
2004-07-02 01:38:04 +08:00
int main(void)
2007-03-09 00:57:36 +08:00
{
2004-07-02 01:38:04 +08:00
/* SZ_encoder_enabled returns 1 if encoder is present */
if(SZ_encoder_enabled() == 1)
exit(0);
else
exit(1);
}
], [hdf5_cv_szlib_can_encode=yes], [hdf5_cv_szlib_can_encode=no],)])
2007-03-09 00:57:36 +08:00
2008-10-25 01:19:34 +08:00
AC_DEFINE(HAVE_FILTER_SZIP, 1,
[Define if support for szip filter is enabled])
USE_FILTER_SZIP="yes"
2007-03-09 00:57:36 +08:00
2004-07-02 01:38:04 +08:00
if test ${hdf5_cv_szlib_can_encode} = "yes"; then
AC_MSG_RESULT([yes])
2007-03-09 00:57:36 +08:00
fi
if test ${hdf5_cv_szlib_can_encode} = "no"; then
2004-07-02 01:38:04 +08:00
AC_MSG_RESULT([no])
2007-03-09 00:57:36 +08:00
fi
2003-04-05 01:54:25 +08:00
dnl Add "szip" to external filter list
[svn-r9493] Purpose:
Fix SZIP filter to dynmically detect encoder.
Description:
Solution:
See:
http://hdf.ncsa.uiuc.edu/RFC/SZIP/Szip_dynamic_12_Oct.pdf
Platforms tested:
verbena, arabica, cobalt.
Note RE testing:
All automatic tests work as before.
The feature is a configuration thing, so the tests are manual.
Tested with the following configs:
pre-release version of SZIP, with dynamic library,
1. no SZIP
2. SZIP, decoder only
3. SZIP, encoder+decoder
released (static lib) SZIP
4. no SZIP
5. SZIP, decoder only
6. SZIP, encoder+decoder
after build, switch szip in LD_LIBRARY_PATH at run time
7. SZIP decoder
8. SZIP encoder
Misc. update:
Manifest updated
2004-11-03 03:10:24 +08:00
if test ${hdf5_cv_szlib_can_encode} = "yes"; then
2007-03-09 00:57:36 +08:00
if test "X$EXTERNAL_FILTERS" != "X"; then
EXTERNAL_FILTERS="${EXTERNAL_FILTERS},"
fi
EXTERNAL_FILTERS="${EXTERNAL_FILTERS}szip(encoder)"
fi
if test ${hdf5_cv_szlib_can_encode} = "no"; then
if test "X$EXTERNAL_FILTERS" != "X"; then
EXTERNAL_FILTERS="${EXTERNAL_FILTERS},"
fi
EXTERNAL_FILTERS="${EXTERNAL_FILTERS}szip(no encoder)"
[svn-r9493] Purpose:
Fix SZIP filter to dynmically detect encoder.
Description:
Solution:
See:
http://hdf.ncsa.uiuc.edu/RFC/SZIP/Szip_dynamic_12_Oct.pdf
Platforms tested:
verbena, arabica, cobalt.
Note RE testing:
All automatic tests work as before.
The feature is a configuration thing, so the tests are manual.
Tested with the following configs:
pre-release version of SZIP, with dynamic library,
1. no SZIP
2. SZIP, decoder only
3. SZIP, encoder+decoder
released (static lib) SZIP
4. no SZIP
5. SZIP, decoder only
6. SZIP, encoder+decoder
after build, switch szip in LD_LIBRARY_PATH at run time
7. SZIP decoder
8. SZIP encoder
Misc. update:
Manifest updated
2004-11-03 03:10:24 +08:00
fi
2007-03-09 00:57:36 +08:00
[svn-r6549] Purpose:
This is the first check-in message related to szip support at HDF5 1.6 release
Description:
szip compression support is required by NASA ESDIS. The compression algorithm
is a good compression algorithm for scientific data. In HDF5, we add another filter
function to make szip as a default compression package as we did for gzip(or zlib).
Solution:
In the configure.in and configure, I handled szip exactly the same as what zlib was built into
the HDF5.
Currently if you don't specify in the configure command line with:
./configure --with-szlib=/.....
Unless you put your szip library in your path,you will build the same HDF5 library like before.
Also, I only use static szip library to test with HDF5.
Platforms tested:
Since there are changes of configure.in and configure,I didn't use h5committest.
I tested with four platforms.
1) Linux 2.4 (eirene)
2) Solaris 2.7(arabica)
3) windows 2000(VS 6.0)
4) SGI IRIX6.5-64(modi4)
For test 1)-3), only basic C tests were done
For modi4 test, I tested 64-bit C,parallel and fortran.
ALl tests passed, except a warning message from szip library when checksum is used in some order, which doesn't cause any real problems.
Misc. update:
2003-04-01 23:18:40 +08:00
fi
2011-04-27 02:41:01 +08:00
AM_CONDITIONAL([BUILD_SHARED_SZIP_CONDITIONAL], [test "X$USE_FILTER_SZIP" = "Xyes" && test "X$LL_PATH" != "X"])
2008-10-25 01:19:34 +08:00
2000-04-21 02:41:51 +08:00
dnl Checkpoint the cache
AC_CACHE_SAVE
1999-12-21 04:23:33 +08:00
dnl ----------------------------------------------------------------------
dnl Is the Pthreads library present? It has a header file `pthread.h' and
dnl a library `-lpthread' and their locations might be specified with the
dnl `--with-pthread' command-line switch. The value is an include path
dnl and/or a library path. If the library path is specified then it must
dnl be preceded by a comma.
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([PTHREAD]) PTHREAD=yes
2002-07-18 00:10:46 +08:00
AC_ARG_WITH([pthread],
[AC_HELP_STRING([--with-pthread=DIR],
[Use the Pthreads library [default=no]])],,
withval=no)
2000-09-15 06:19:24 +08:00
case "$withval" in
2001-09-28 07:05:16 +08:00
yes)
2003-07-29 05:38:04 +08:00
AC_CHECK_HEADERS([pthread.h])
[svn-r20657] Description:
Correct check for pthreads routine from pthread_join() to pthread_self()
as intended.
Tested on
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
FreeBSD/32 8.2 (loyalty) w/gcc4.6, w/C++ & FORTRAN, in debug mode
FreeBSD/64 8.2 (freedom) w/gcc4.6, w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (koala) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, w/threadsafe, in production mode
Linux/PPC 2.6 (heiwa) w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-ia64 2.6 (ember) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in debug mode
2011-04-28 08:56:57 +08:00
AC_CHECK_LIB([pthread], [pthread_self],, [unset PTHREAD])
2001-09-28 07:05:16 +08:00
;;
no)
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for pthread])
AC_MSG_RESULT([suppressed])
2001-09-28 07:05:16 +08:00
unset PTHREAD
;;
*)
2001-12-18 02:00:24 +08:00
case "$withval" in
*,*)
pthread_inc="`echo $withval | cut -f1 -d,`"
pthread_lib="`echo $withval | cut -f2 -d, -s`"
;;
*)
if test -n "$withval"; then
pthread_inc="$withval/include"
pthread_lib="$withval/lib"
fi
;;
esac
2003-01-24 03:19:02 +08:00
dnl Trying to include -I/usr/include and -L/usr/lib is redundant and
dnl can mess some compilers up.
if test "X$pthread_inc" = "X/usr/include"; then
pthread_inc=""
fi
if test "X$pthread_lib" = "X/usr/lib"; then
pthread_lib=""
fi
2001-09-28 07:05:16 +08:00
if test -n "$pthread_inc"; then
saved_CPPFLAGS="$CPPFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_CPPFLAGS="$AM_CPPFLAGS"
2001-09-28 07:05:16 +08:00
CPPFLAGS="$CPPFLAGS -I$pthread_inc"
2009-10-15 05:43:05 +08:00
AM_CPPFLAGS="$AM_CPPFLAGS -I$pthread_inc"
AC_CHECK_HEADERS([pthread.h],, [CPPFLAGS="$saved_CPPFLAGS"; AM_CPPFLAGS="$saved_AM_CPPFLAGS"; unset PTHREAD])
2001-09-28 07:05:16 +08:00
else
2003-07-29 05:38:04 +08:00
AC_CHECK_HEADERS([pthread.h],, [unset PTHREAD])
2001-09-28 07:05:16 +08:00
fi
2000-09-15 06:19:24 +08:00
2001-09-28 07:05:16 +08:00
if test -n "$pthread_lib"; then
saved_LDFLAGS="$LDFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_LDFLAGS="$AM_LDFLAGS"
2001-09-28 07:05:16 +08:00
LDFLAGS="$LDFLAGS -L$pthread_lib"
2009-10-15 05:43:05 +08:00
AM_LDFLAGS="$AM_LDFLAGS -L$pthread_lib"
[svn-r20657] Description:
Correct check for pthreads routine from pthread_join() to pthread_self()
as intended.
Tested on
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
FreeBSD/32 8.2 (loyalty) w/gcc4.6, w/C++ & FORTRAN, in debug mode
FreeBSD/64 8.2 (freedom) w/gcc4.6, w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (koala) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, w/threadsafe, in production mode
Linux/PPC 2.6 (heiwa) w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-ia64 2.6 (ember) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in debug mode
2011-04-28 08:56:57 +08:00
AC_CHECK_LIB([pthread], [pthread_self],,
2009-10-15 05:43:05 +08:00
[LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset PTHREAD])
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2001-09-28 07:05:16 +08:00
else
[svn-r20657] Description:
Correct check for pthreads routine from pthread_join() to pthread_self()
as intended.
Tested on
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
FreeBSD/32 8.2 (loyalty) w/gcc4.6, w/C++ & FORTRAN, in debug mode
FreeBSD/64 8.2 (freedom) w/gcc4.6, w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (koala) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, w/threadsafe, in production mode
Linux/PPC 2.6 (heiwa) w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-ia64 2.6 (ember) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in debug mode
2011-04-28 08:56:57 +08:00
AC_CHECK_LIB([pthread], [pthread_self],, [unset PTHREAD])
2001-09-28 07:05:16 +08:00
fi
;;
1999-12-21 04:23:33 +08:00
esac
2000-06-06 07:25:29 +08:00
2000-04-11 02:23:38 +08:00
dnl ----------------------------------------------------------------------
2002-01-03 03:05:30 +08:00
dnl Enable thread-safe version of library. It requires Pthreads support.
2000-04-11 02:23:38 +08:00
dnl
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for thread safe support])
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([threadsafe],
[AC_HELP_STRING([--enable-threadsafe],
[Enable thread safe capability])],
THREADSAFE=$enableval)
2000-09-15 06:19:24 +08:00
case "X-$THREADSAFE" in
2001-09-18 06:47:02 +08:00
X-|X-no)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([no])
2001-09-18 06:47:02 +08:00
;;
X-yes)
2002-01-03 03:05:30 +08:00
dnl Check that we can link a simple Pthread program.
[svn-r20657] Description:
Correct check for pthreads routine from pthread_join() to pthread_self()
as intended.
Tested on
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
FreeBSD/32 8.2 (loyalty) w/gcc4.6, w/C++ & FORTRAN, in debug mode
FreeBSD/64 8.2 (freedom) w/gcc4.6, w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (koala) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, w/threadsafe, in production mode
Linux/PPC 2.6 (heiwa) w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-ia64 2.6 (ember) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in debug mode
2011-04-28 08:56:57 +08:00
AC_TRY_LINK(, [pthread_self()],
2003-07-29 05:38:04 +08:00
[AC_MSG_RESULT([yes]); THREADSAFE=yes],
[AC_MSG_ERROR([needed pthread library not available])])
2001-09-18 06:47:02 +08:00
;;
*)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([error])
AC_MSG_ERROR([\'$enableval\' is not a valid threadsafe type])
2001-09-18 06:47:02 +08:00
;;
2000-04-11 02:23:38 +08:00
esac
2000-06-06 07:25:29 +08:00
2000-09-15 06:19:24 +08:00
if test "X$THREADSAFE" = "Xyes"; then
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_THREADSAFE], [1], [Define if we have thread safe support])
2000-04-11 02:23:38 +08:00
fi
1999-12-21 04:23:33 +08:00
1998-07-30 00:43:59 +08:00
dnl ----------------------------------------------------------------------
dnl How does one figure out the local time zone? Anyone know of a
dnl Posix way to do this?
dnl
dnl First check if `struct tm' has a `tm_gmtoff' member.
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for tm_gmtoff in struct tm])
1998-07-30 00:43:59 +08:00
AC_TRY_COMPILE([
#include <sys/time.h>
2003-07-29 05:38:04 +08:00
#include <time.h>], [struct tm tm; tm.tm_gmtoff=0;],
AC_DEFINE([HAVE_TM_GMTOFF], [1],
2005-02-01 11:17:02 +08:00
[Define if `tm_gmtoff' is a member of `struct tm'])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
1998-07-30 00:43:59 +08:00
1999-02-16 01:38:04 +08:00
dnl check if `struct tm' has a `__tm_gmtoff' member.
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for __tm_gmtoff in struct tm])
1999-02-16 01:38:04 +08:00
AC_TRY_COMPILE([
#include <sys/time.h>
2003-07-29 05:38:04 +08:00
#include <time.h>], [struct tm tm; tm.__tm_gmtoff=0;],
AC_DEFINE([HAVE___TM_GMTOFF], [1],
2005-02-01 11:17:02 +08:00
[Define if `__tm_gmtoff' is a member of `struct tm'])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
1999-02-16 01:38:04 +08:00
1998-07-30 00:43:59 +08:00
dnl Check whether the global variable `timezone' is defined.
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for global timezone variable])
2008-05-01 03:51:13 +08:00
case "`uname`" in
CYGWIN*)
AC_MSG_RESULT([disabled in CYGWIN])
;;
*)
AC_TRY_LINK([
#include <sys/time.h>
#include <time.h>], [timezone=0;],
AC_DEFINE([HAVE_TIMEZONE], [1],
[Define if `timezone' is a global variable])
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
;;
esac
1998-07-30 00:43:59 +08:00
dnl Check whether `struct timezone' is defined.
AC_STRUCT_TIMEZONE
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for struct timezone])
1998-07-30 00:43:59 +08:00
AC_TRY_COMPILE([
#include <sys/types.h>
#include <sys/time.h>
2003-07-29 05:38:04 +08:00
#include <time.h>], [struct timezone tz; tz.tz_minuteswest=0;],
AC_DEFINE([HAVE_STRUCT_TIMEZONE], [1],
2005-02-01 11:17:02 +08:00
[Define if `struct timezone' is defined])
2004-07-24 01:29:16 +08:00
have_struct_tz="yes"
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
1998-07-30 00:43:59 +08:00
2004-07-24 01:29:16 +08:00
dnl If gettimeofday() is going to be used, make sure it uses the timezone struct
if test "$have_gettime" = "yes" -a "$have_struct_tz" = "yes"; then
AC_MSG_CHECKING(whether gettimeofday() gives timezone)
2004-08-11 00:53:26 +08:00
AC_CACHE_VAL([hdf5_cv_gettimeofday_tz],
[AC_TRY_RUN([
#include <time.h>
#include <sys/time.h>
int main(void)
{
struct timeval tv;
struct timezone tz;
tz.tz_minuteswest = 7777; /* Initialize to an unreasonable number */
tz.tz_dsttime = 7;
gettimeofday(&tv, &tz);
/* Check whether the function returned any value at all */
if(tz.tz_minuteswest == 7777 && tz.tz_dsttime == 7)
exit(1);
else exit (0);
}], [hdf5_cv_gettimeofday_tz=yes], [hdf5_cv_gettimeofday_tz=no],)])
if test ${hdf5_cv_gettimeofday_tz} = "yes"; then
2004-07-24 01:29:16 +08:00
AC_MSG_RESULT([yes])
AC_DEFINE([GETTIMEOFDAY_GIVES_TZ], [1],
[Define if gettimeofday() populates the tz pointer passed in])
else
AC_MSG_RESULT([no])
fi
fi
1998-11-03 01:58:28 +08:00
dnl ----------------------------------------------------------------------
dnl Does the struct stat have the st_blocks field? This field is not Posix.
dnl
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for st_blocks in struct stat])
1998-11-03 01:58:28 +08:00
AC_TRY_COMPILE([
#include <sys/stat.h>],[struct stat sb; sb.st_blocks=0;],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_STAT_ST_BLOCKS], [1],
2005-02-01 11:17:02 +08:00
[Define if `struct stat' has the `st_blocks' field])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
[svn-r339] Changes since 19980408
----------------------
./src/H5Osdspace.c
./html/H5.format.html
In the past we were allowed to have >2GB files on a 32-bit
machine as long as no dataset within the file was larger than
4GB (or whatever sizeof(size_t) is). That's been fixed now.
All dataset size calculations are done with `hsize_t' which is
normally defined as `unsigned long long'.
./src/H5F.c
./src/H5Ffamily.c
./src/H5Fprivate.h
./src/H5P.c
./src/H5Ppublic.h
The file family member size can now be set/queried. The
default is still 64MB, but it can be set to 1GB by saying:
H5Pset_family (plist, 30, H5P_DEFAULT);
When opening an existing file family the specified
bits-per-member is ignored and the first member of the family
determines the bits-per-member, which can be retrieved with
H5Pget_family().
./acconfig.h
./configure.in
./src/H5config.h
./src/H5public.h
Added `--disable-hsizet' so that those with old GCC compilers
(<2.8.1) can still compile the code.
./src/H5.c
./src/H5private.h
Added HDfprintf() which works just like fprintf() except you
can give `H' as a size modifier for the integer conversions
and supply an `hsize_t' or `hssize_t' argument without casting
it. For instance:
hsize_t npoints = H5Sget_npoints(space);
HDfprintf(stdout,"Dataset has %Hd (%#018Hx) points\n",
npoints, npoints);
You can now give `%a' as a format to print an address, but all
formating flags are ignored and it causes the return value of
HDfprintf() to not include the characters in the address (but
who uses the return value anyway :-). Example:
H5G_t *grp;
HDfprintf(stdout, "Group object header at %a\n",
&(grp->ent.header));
Added HDstrtoll() which works exactly like [HD]strtol() except
the result is an int64.
./src/debug.c
Large addresses can now be entered from the command-line. Use
either decimal, octal (leading `0') or hexadecimal (leading
`0x') when giving the address.
./src/h5ls.c
The printf format for dataset dimensions was changed to `%Hu'
to support large datasets.
./test/big.c [NEW]
A test for big datasets on 32-bit machines. This test is not
run by default. Don't try to run it on an nfs-mounted file
system or other file system that doesn't support holes because
it creates two 32GB datasets of all zero.
1998-04-10 04:22:11 +08:00
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
dnl ----------------------------------------------------------------------
dnl How do we figure out the width of a tty in characters?
dnl
AC_CHECK_FUNCS(_getvideoconfig gettextinfo GetConsoleScreenBufferInfo)
1999-03-18 21:42:52 +08:00
AC_CHECK_FUNCS(_scrsize ioctl)
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for struct videoconfig])
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
AC_TRY_COMPILE(,[struct videoconfig w; w.numtextcols=0;],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_STRUCT_VIDEOCONFIG], [1],
2005-02-01 11:17:02 +08:00
[Define if `struct videoconfig' is defined])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for struct text_info])
AC_TRY_COMPILE(, [struct text_info w; w.screenwidth=0;],
AC_DEFINE([HAVE_STRUCT_TEXT_INFO], [1],
2005-02-01 11:17:02 +08:00
[Define if `struct text_info' is defined])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for TIOCGWINSZ])
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
AC_TRY_COMPILE([#include <sys/ioctl.h>],[int w=TIOCGWINSZ;],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_TIOCGWINSZ], [1],
2002-04-03 06:08:23 +08:00
[Define if the ioctl TIOGWINSZ is defined])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
2003-08-09 02:28:03 +08:00
AC_MSG_CHECKING([for TIOCGETD])
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
AC_TRY_COMPILE([#include <sys/ioctl.h>],[int w=TIOCGETD;],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_TIOCGETD], [1],
2002-04-03 06:08:23 +08:00
[Define if the ioctl TIOCGETD is defined])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
[svn-r879] Changes since 19981105
----------------------
./configure.in
./acconfig.h
./configure [REGENERATED]
./src/H5config.h.in [REGENERATED]
Added checks for functions, structs, constants, and header
files used in h5ls.c to determine the output width with
various Windows compilers and Unix variants.
Added check for <sys/stat.h> and defined
HAVE_SYS_STAT_H. Included <sys/stat.h> in H5private.h outside
the Posix.1 #include's section.
./src/H5RA.c
Less aggressive about failing -- rather returns false, which
allows type detection to continue.
./src/h5ls.c
Data types are displayed for datasets and named data types when
`-v' or `--verbose' is specified on the command-line. The
algorithm is a little different than the dumper because we're
trying to be human-friendly, not necessarily machine-friendly.
* Any data type which matches a native C type gets printed
something like `native double'.
* A floating point type that matches one of the IEEE standard
types but not one of the native types gets printed like `IEEE
64-bit big-endian float'.
* Other floating point values have information about sign bit
location; exponent size, location, and bias; and significand
size, location, and normalization.
* Padding and offsets are displayed for types that have
padding (precision != size), including internal padding for
some floating point data types.
* Non-native integer types are displayed like `32-bit
little-endian unsigned integer'.
* Compound data types have each member displayed including the
member name, byte offset within the struct, dimensions, index
permutation, and data type.
* String types are displayed like `256-byte null-terminated
ASCII string'.
* References are displayed like `8-byte unknown reference'
until the reference interface stabilizes a little.
* All other types including types not yet defined will be
printed like `4-byte class-9 unknown'.
The dimensionality of scalar datasets is printed like `{SCALAR}'
instead of just `{}'.
If external raw files are used to store a dataset then the offsets,
sizes, and file names of each are printed if `-v' or `--verbose'
was given on the command-line.
If an object is found and h5ls can't determine the object type then
it still tries to print the number of hard links, the OID, and any
comment that might be present if `-v' or `--verbose' was specified.
If the `-d' or `--dump' switch is turned on then ragged arrays will
report that the data can only be dumped by dumping the component
datasets explicitly. I'm not planning to implement this since
we're going to eventually change the whole way ragged arrays are
stored.
Compound data values do not have the component names displayed by
default when `-v' or `--verbose' is turned on. Instead, the names
can be displayed with `-l' or `--label'.
The output width is determined by the first rule that applies:
* If the `-wN', `-w N' or `--width=N' switch appeared on
the command line then use N for the output width.
* Query the OS for the tty width in a highly unportable way
borrowed from GNU `less' depending on what functions and
data structures were found during configuration (if any):
_getvideoconfig(), gettextinfo(), _srcsize(), ioctl(),
GetConsoleScreenBufferInfo(), struct videoconfig, struct
text_info, the TIOCGWINSZ ioctl, the TIOCGETD ioctl.
* If the `COLUMNS' environment variable is set then use
its value.
* Use the value 80.
Just for kicks, run Mark and Jim's test_vbt and then say `h5ls -dlsv
test.vbt'. You can also try it on the various *.h5 files in the
test/example directories.
./config/linux
Removed turning on parallel by default on Robb's macine.
1998-11-07 02:00:22 +08:00
[svn-r339] Changes since 19980408
----------------------
./src/H5Osdspace.c
./html/H5.format.html
In the past we were allowed to have >2GB files on a 32-bit
machine as long as no dataset within the file was larger than
4GB (or whatever sizeof(size_t) is). That's been fixed now.
All dataset size calculations are done with `hsize_t' which is
normally defined as `unsigned long long'.
./src/H5F.c
./src/H5Ffamily.c
./src/H5Fprivate.h
./src/H5P.c
./src/H5Ppublic.h
The file family member size can now be set/queried. The
default is still 64MB, but it can be set to 1GB by saying:
H5Pset_family (plist, 30, H5P_DEFAULT);
When opening an existing file family the specified
bits-per-member is ignored and the first member of the family
determines the bits-per-member, which can be retrieved with
H5Pget_family().
./acconfig.h
./configure.in
./src/H5config.h
./src/H5public.h
Added `--disable-hsizet' so that those with old GCC compilers
(<2.8.1) can still compile the code.
./src/H5.c
./src/H5private.h
Added HDfprintf() which works just like fprintf() except you
can give `H' as a size modifier for the integer conversions
and supply an `hsize_t' or `hssize_t' argument without casting
it. For instance:
hsize_t npoints = H5Sget_npoints(space);
HDfprintf(stdout,"Dataset has %Hd (%#018Hx) points\n",
npoints, npoints);
You can now give `%a' as a format to print an address, but all
formating flags are ignored and it causes the return value of
HDfprintf() to not include the characters in the address (but
who uses the return value anyway :-). Example:
H5G_t *grp;
HDfprintf(stdout, "Group object header at %a\n",
&(grp->ent.header));
Added HDstrtoll() which works exactly like [HD]strtol() except
the result is an int64.
./src/debug.c
Large addresses can now be entered from the command-line. Use
either decimal, octal (leading `0') or hexadecimal (leading
`0x') when giving the address.
./src/h5ls.c
The printf format for dataset dimensions was changed to `%Hu'
to support large datasets.
./test/big.c [NEW]
A test for big datasets on 32-bit machines. This test is not
run by default. Don't try to run it on an nfs-mounted file
system or other file system that doesn't support holes because
it creates two 32GB datasets of all zero.
1998-04-10 04:22:11 +08:00
dnl ----------------------------------------------------------------------
dnl Check for functions.
dnl
[svn-r17924] Description:
Add detection for POSIX lstat() routine to configure script (mostly for
non-UNIX/Linux machines) and add macro wrapper for it.
Alphabetatize the system/library calls we test for, to make them easier
to read.
Removed the sigaction() detection & macro wrappers, since it's not
used by the distribution currently.
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.6.2 (amazon) in debug mode
Mac OS X/32 10.6.2 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2009-11-24 05:34:51 +08:00
AC_CHECK_FUNCS(alarm BSDgettimeofday fork frexpf frexpl)
2010-04-29 03:56:29 +08:00
AC_CHECK_FUNCS(gethostname getpwuid getrusage lstat)
AC_CHECK_FUNCS(rand_r random setsysinfo)
AC_CHECK_FUNCS(signal longjmp setjmp siglongjmp sigsetjmp sigprocmask)
[svn-r17939] Description:
Handle external links from symlinked files by adding another check to
look for "child" files for links from the actual location of the "parent" file,
instead of from the location of the symlink.
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.6.2 (amazon) in debug mode
Mac OS X/32 10.6.2 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2009-12-01 21:39:05 +08:00
AC_CHECK_FUNCS(snprintf srandom strdup symlink system)
AC_CHECK_FUNCS(tmpfile vasprintf waitpid)
2005-01-24 11:26:37 +08:00
dnl Check for vsnprintf() separately, so we can detect situations where it
dnl doesn't return the correct size for formatted strings that are too large
dnl for the buffer provided
AC_CHECK_FUNCS(vsnprintf,
dnl Check if vsnprintf() returns correct size for strings that don't fit
dnl into the size allowed. If vsnprintf() works correctly on this platform,
2005-01-25 02:50:31 +08:00
dnl it should return a value of 42 for the test below
dnl
dnl Note that vsnprintf fails in two different ways:
dnl - In IRIX64, calls to vnsprintf() with a formatted string that
dnl is larger than the buffer size allowed incorrectly
dnl return the size of the buffer minus one.
dnl - In HP/UX, calls to vsnprintf() with a formatted string that
dnl is larger than the buffer size allowed incorrectly
dnl return (-1)
2005-01-24 11:26:37 +08:00
AC_MSG_CHECKING([if vsnprintf returns correct value])
AC_CACHE_VAL([hdf5_cv_vsnprintf_works],
AC_TRY_RUN([
#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
int test_vsnprintf(const char *fmt,...)
{
va_list ap;
char *s = malloc(16);
int ret;
va_start(ap, fmt);
ret=vsnprintf(s,16,"%s",ap);
va_end(ap);
2005-01-25 02:50:31 +08:00
return(ret!=42 ? 1 : 0);
2005-01-24 11:26:37 +08:00
}
int main(void)
{
exit(test_vsnprintf("%s","A string that is longer than 16 characters"));
}
],[hdf5_cv_vsnprintf_works=yes],[hdf5_cv_vsnprintf_works=no],))
if test ${hdf5_cv_vsnprintf_works} = "yes"; then
AC_MSG_RESULT([yes])
AC_DEFINE([VSNPRINTF_WORKS], [1],
[Define if vsnprintf() returns the correct value for formatted strings that don't fit into size allowed])
else
AC_MSG_RESULT([no])
fi
,)
1997-08-16 00:35:53 +08:00
2006-06-14 05:21:14 +08:00
dnl ----------------------------------------------------------------------
dnl Check that a lone colon can be used as an argument
dnl This is not true on Cray X1, which interprets a lone colon as a
dnl system command.
dnl
AC_CACHE_CHECK([if lone colon can be used as an argument],
[hdf5_cv_lone_colon],
[
echo "int main(int argc, char * argv[]) {return 0;}" > conftest.c
$CC $CFLAGS conftest.c > /dev/null 2> /dev/null
echo "./a.out :" > conftest.sh
chmod 700 conftest.sh
./conftest.sh 2> conftest.out
2007-02-07 02:50:04 +08:00
rm a.out
2006-06-14 05:21:14 +08:00
TEST_OUTPUT=`cat conftest.out`
if test "X$TEST_OUTPUT" = "X"; then
hdf5_cv_lone_colon=yes
else
hdf5_cv_lone_colon=no
fi
])
AC_SUBST(H5_LONE_COLON) H5_LONE_COLON="$hdf5_cv_lone_colon"
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
1997-10-21 07:14:35 +08:00
dnl Check compiler characteristics
dnl
AC_C_CONST
AC_C_INLINE
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for __attribute__ extension])
1999-03-04 07:17:48 +08:00
AC_TRY_COMPILE(,[int __attribute__((unused)) x],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_ATTRIBUTE], [1],
2002-04-03 06:08:23 +08:00
[Define if the __attribute__(()) extension is present])
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
1997-10-21 07:14:35 +08:00
2008-09-17 22:04:35 +08:00
AC_MSG_CHECKING([for __func__ extension])
AC_TRY_COMPILE(,[ const char *fname = __func__; ],
AC_DEFINE([HAVE_C99_FUNC], [1],
2009-01-14 21:09:42 +08:00
[Define if the compiler understands the __func__ keyword])
2008-09-17 22:04:35 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for __FUNCTION__ extension])
2008-09-17 22:04:35 +08:00
AC_TRY_COMPILE(,[ const char *fname = __FUNCTION__; ],
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_FUNCTION], [1],
2009-01-14 21:09:42 +08:00
[Define if the compiler understands the __FUNCTION__ keyword])
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
AC_MSG_CHECKING([for C99 designated initialization support])
AC_TRY_COMPILE(,[
typedef struct {
int x;
union {
int i;
double d;
} u;
} di_struct_t;
di_struct_t x = {0, { .d = 0.0}}; ],
AC_DEFINE([HAVE_C99_DESIGNATED_INITIALIZER], [1],
[Define if the compiler understands C99 designated initialization of structs and unions])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no]))
1997-08-16 00:35:53 +08:00
1998-04-25 05:26:32 +08:00
dnl ----------------------------------------------------------------------
dnl Try to figure out how to print `long long'. Some machines use `%lld'
dnl and others use `%qd'. There may be more! The final `l' is a
dnl default in case none of the others work.
2000-02-18 20:53:32 +08:00
dnl Need to patch up LD_LIBRARY_PATH so that the execution can find all
dnl the dynamic library. The correct way to do it should be updating
dnl LD_LIBRARY_PATH along with LDFLAGS or do it with the AC_TRY_RUN macro.
1998-04-25 05:26:32 +08:00
dnl
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([how to print long long])
2002-07-18 00:10:46 +08:00
AC_CACHE_VAL([hdf5_cv_printf_ll],
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
LD_LIBRARY_PATH="$LD_LIBRARY_PATH`echo $AM_LDFLAGS $LDFLAGS | sed -e 's/-L/:/g' -e 's/ //g'`"
2000-02-18 20:53:32 +08:00
export LD_LIBRARY_PATH
2000-09-15 06:19:24 +08:00
2004-12-29 22:26:20 +08:00
for hdf5_cv_printf_ll in l ll L q unknown; do
2000-09-15 06:19:24 +08:00
AC_TRY_RUN([
1998-04-25 05:26:32 +08:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
2000-09-15 06:19:24 +08:00
int main(void)
{
2001-12-18 02:00:24 +08:00
char *s = malloc(128);
2000-09-15 06:19:24 +08:00
long long x = (long long)1048576 * (long long)1048576;
sprintf(s,"%${hdf5_cv_printf_ll}d",x);
exit(strcmp(s,"1099511627776"));
}
2001-07-18 05:24:38 +08:00
], break,,continue)
2002-07-18 00:10:46 +08:00
done)dnl
2000-09-15 06:19:24 +08:00
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([%${hdf5_cv_printf_ll}d and %${hdf5_cv_printf_ll}u])
AC_DEFINE_UNQUOTED([PRINTF_LL_WIDTH], ["$hdf5_cv_printf_ll"],
2005-02-01 11:17:02 +08:00
[Width for printf() for type `long long' or `__int64', use `ll'])
1998-04-25 05:26:32 +08:00
2003-02-10 23:38:52 +08:00
dnl ----------------------------------------------------------------------
dnl Check if pthread_attr_setscope(&attribute, PTHREAD_SCOPE_SYSTEM)
dnl is supported on this system
dnl
AC_MSG_CHECKING([Threads support system scope])
2003-07-29 05:38:04 +08:00
AC_CACHE_VAL([hdf5_cv_system_scope_threads],
[AC_TRY_RUN([
2003-02-10 23:38:52 +08:00
#if STDC_HEADERS
#include <stdlib.h>
#include <pthread.h>
#endif
int main(void)
{
pthread_attr_t attribute;
int ret;
pthread_attr_init(&attribute);
ret=pthread_attr_setscope(&attribute, PTHREAD_SCOPE_SYSTEM);
exit(ret==0 ? 0 : 1);
}
2003-07-29 05:38:04 +08:00
], [hdf5_cv_system_scope_threads=yes], [hdf5_cv_system_scope_threads=no],)])
2003-02-10 23:38:52 +08:00
if test ${hdf5_cv_system_scope_threads} = "yes"; then
2003-07-29 05:38:04 +08:00
AC_DEFINE([SYSTEM_SCOPE_THREADS], [1],
2003-02-10 23:38:52 +08:00
[Define if your system supports pthread_attr_setscope(&attribute, PTHREAD_SCOPE_SYSTEM) call.])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes])
2003-02-10 23:38:52 +08:00
else
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([no])
2003-02-10 23:38:52 +08:00
fi
2002-07-18 00:10:46 +08:00
dnl ----------------------------------------------------------------------
dnl Checking to see if GPFS is available on this filesystem
dnl
2003-05-18 03:38:50 +08:00
AC_ARG_ENABLE([gpfs],
[AC_HELP_STRING([--enable-gpfs],
[Enable GPFS hints for the MPI/POSIX file
driver. [default=no]])],,
[enableval=no])
case "X-$enableval" in
X-yes)
AC_CHECK_HEADERS([gpfs.h],
AC_MSG_CHECKING([for GPFS support])
AC_TRY_COMPILE([#include <gpfs.h>],
[int fd = 0; gpfs_fcntl(fd, (void *)0);],
AC_DEFINE(HAVE_GPFS, 1,
[Define if we have GPFS support])
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([yes])
2003-05-18 03:38:50 +08:00
LIBS="$LIBS -lgpfs"
GPFS="yes",
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([no])
2003-05-18 03:38:50 +08:00
GPFS="no"))
;;
X-no|*)
AC_MSG_CHECKING([for gpfs])
AC_MSG_RESULT([suppressed])
;;
esac
2002-07-18 00:10:46 +08:00
1998-04-03 11:29:38 +08:00
dnl ----------------------------------------------------------------------
dnl Turn on debugging by setting compiler flags
2007-04-27 22:56:33 +08:00
dnl This must come after the enable-production since it depends on production.
1998-04-03 11:29:38 +08:00
dnl
[svn-r429] Changes since 19980616
----------------------
./html/tracing.html NEW
This entire update is to make it possible for the library to
print the name, arguments, and return value of every API call
without requiring any extra work from developers or app
programmers. This file describes how this all works.
./configure.in
Added the `--enable-tracing' switch. If you use it then the
library will include code to print API function names,
argument names and values, and function return values.
However, you must then turn on the tracing by setting the
HDF5_TRACE environment variable to a file descriptor number.
The default is `--disable-tracing' since enabling it causes a
slight increase in library size and a slowdown resulting from
an extra function call for each API function call (I couldn't
even measure the slowdown :-)
./bin/trace NEW
A perl script that synchronizes the H5TRACE() macro calls in
the *.c files with the function return type and formal
argument names and types. If you use GNU make and gcc then
this will be done automatically, otherwise just invoke this
script with the names of one or more .c files. You could do
it by hand to, but encoding argument types is a little tricky
at first.
./config/commence.in
Added the $(TRACE) macro, which defaults to the no-op.
Added -D_POSIX_SOURCE to the compiler command line.
./src/Makefile.in
Override the default for $(TRACE).
./config/depend.in
Automatically calls $(TRACE) to synchronize the H5TRACE()
macros in any source file that changed. As with makefile
dependencies, one way to force synchronization of all files is
to remove the `.depend' file.
./MANIFEST
Added new files.
./src/H5Eprivate.h
Modified HRETURN_ERROR() and HRETURN() for tracing.
./src/H5.c
./src/H5private.h
This is where the real tracing work really happens, in
H5_trace().
./src/H5A.c
./src/H5D.c
./src/H5G.c
./src/H5P.c
./src/H5S.c
./src/H5Z.c
Added H5TRACE() calls to all API functions. You don't really
need these changes if you don't want to merge your stuff
because they can be generated automatically by going to the
hdf5/src directory and saying ../bin/trace *.c
./src/H5T.c
Added H5TRACE() calls. Other stuff below.
./src/H5E.c
./src/H5Epublic.h
Added H5TRACE() calls. Created a type H5E_auto_t for the
`func' argument of H5Eset_auto() and H5Eget_auto() to make
those arguments easier to parse for tracing. It should also
make it clearer for users that don't know how to read
complicated ANSI data types.
./src/H5F.c
Added H5TRACE() calls. Changed a couple `uintn' argument
types in API functions to `unsigned int' since `uintn' part of
the API. Changed a few "can't" and "cant" error messages to
"unable to".
./src/H5Ipublic.h
Removed H5_DIRECTORY from the H5I_group_t enum. It wasn't
used anywhere.
./src/H5Tconv.c
Removed an unused label.
./src/H5Fistore.c
./src/H5Oattr.c
./src/H5Odtype.c
./src/H5T.c
./test/dsets.c
./test/dtypes.c
Fixed a warning about a variable possibly used before it's
initialized. Added __unused__ to turn off some unused
argument warnings that pop up when debugging is turned off and
optimizations are turned on.
1998-06-18 04:46:29 +08:00
AC_MSG_CHECKING(for debug flags)
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([debug],
[AC_HELP_STRING([--enable-debug=all],
[Turn on debugging in all packages. One may
also specify a comma-separated list of
package names without the leading H5 or
2007-05-01 11:38:52 +08:00
the word no. The default is most packages
if production is disabled; no if it is enabled.
])],
2002-07-18 00:10:46 +08:00
[DEBUG_PKG=$enableval])
2007-04-27 22:56:33 +08:00
dnl Default to no if producton is enabled
if test "X-$DEBUG_PKG" = X- ; then
if test "$enable_production" = yes ; then
DEBUG_PKG=no
else
DEBUG_PKG=yes
fi
fi
2002-07-18 00:10:46 +08:00
AC_SUBST([DEBUG_PKG])
2008-12-16 01:39:20 +08:00
all_packages="ac,b,b2,d,e,f,g,hg,hl,i,mf,mm,o,p,s,t,v,z"
2000-09-15 06:19:24 +08:00
case "X-$DEBUG_PKG" in
2007-04-27 22:56:33 +08:00
X-yes)
2001-09-18 06:47:02 +08:00
DEBUG_PKG="d,e,f,g,hg,i,mm,o,p,s,t,v,z"
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -UNDEBUG"
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([default ($DEBUG_PKG)])
2001-09-18 06:47:02 +08:00
;;
X-all)
DEBUG_PKG=$all_packages
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -UNDEBUG"
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([all ($DEBUG_PKG)])
2001-09-18 06:47:02 +08:00
;;
2007-03-17 07:26:30 +08:00
X-no|X-none)
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([none])
2001-09-18 06:47:02 +08:00
DEBUG_PKG=
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -DNDEBUG"
2001-09-18 06:47:02 +08:00
;;
*)
2003-08-09 02:28:03 +08:00
AC_MSG_RESULT([$DEBUG_PKG])
2001-09-18 06:47:02 +08:00
;;
2000-03-31 05:33:24 +08:00
esac
1998-04-03 11:29:38 +08:00
2000-09-15 06:19:24 +08:00
if test -n "$DEBUG_PKG"; then
2006-10-04 06:41:34 +08:00
for pkg in `echo $DEBUG_PKG | ${TR} ${as_cr_letters}"," ${as_cr_LETTERS}" "`; do
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -DH5${pkg}_DEBUG"
2001-09-18 06:47:02 +08:00
done
1998-04-03 11:29:38 +08:00
fi
1997-08-16 00:35:53 +08:00
2003-06-26 10:08:36 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they would like the function stack support compiled in
dnl
AC_MSG_CHECKING([whether function stack tracking is enabled])
2006-05-02 11:17:49 +08:00
AC_ARG_ENABLE([codestack],
[AC_HELP_STRING([--enable-codestack],
2006-03-23 02:44:55 +08:00
[Enable the function stack tracing (for developer debugging).])],
2006-05-02 11:17:49 +08:00
[CODESTACK=$enableval])
2003-06-26 10:08:36 +08:00
2006-05-02 11:17:49 +08:00
case "X-$CODESTACK" in
2006-03-23 02:44:55 +08:00
X-yes)
2006-05-02 11:17:49 +08:00
CODESTACK=yes
2003-06-26 10:08:36 +08:00
AC_MSG_RESULT([yes])
2006-05-02 11:17:49 +08:00
AC_DEFINE([HAVE_CODESTACK], [1],
2003-06-26 10:08:36 +08:00
[Define if the function stack tracing code is to be compiled in])
;;
*)
2006-05-02 11:17:49 +08:00
CODESTACK=no
2003-06-26 10:08:36 +08:00
AC_MSG_RESULT([no])
;;
esac
2006-07-14 01:59:39 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they would like the metadata trace file code compiled in
dnl
AC_MSG_CHECKING([whether metadata trace file code is enabled])
AC_ARG_ENABLE([metadata-trace-file],
[AC_HELP_STRING([--enable-metadata-trace-file],
[Enable metadata trace file collection.])],
[METADATATRACEFILE=$enableval])
case "X-$METADATATRACEFILE" in
X-yes)
METADATATRACEFILE=yes
AC_MSG_RESULT([yes])
AC_DEFINE([METADATA_TRACE_FILE], [1],
[Define if the metadata trace file code is to be compiled in])
;;
*)
METADATATRACEFILE=no
AC_MSG_RESULT([no])
;;
esac
[svn-r429] Changes since 19980616
----------------------
./html/tracing.html NEW
This entire update is to make it possible for the library to
print the name, arguments, and return value of every API call
without requiring any extra work from developers or app
programmers. This file describes how this all works.
./configure.in
Added the `--enable-tracing' switch. If you use it then the
library will include code to print API function names,
argument names and values, and function return values.
However, you must then turn on the tracing by setting the
HDF5_TRACE environment variable to a file descriptor number.
The default is `--disable-tracing' since enabling it causes a
slight increase in library size and a slowdown resulting from
an extra function call for each API function call (I couldn't
even measure the slowdown :-)
./bin/trace NEW
A perl script that synchronizes the H5TRACE() macro calls in
the *.c files with the function return type and formal
argument names and types. If you use GNU make and gcc then
this will be done automatically, otherwise just invoke this
script with the names of one or more .c files. You could do
it by hand to, but encoding argument types is a little tricky
at first.
./config/commence.in
Added the $(TRACE) macro, which defaults to the no-op.
Added -D_POSIX_SOURCE to the compiler command line.
./src/Makefile.in
Override the default for $(TRACE).
./config/depend.in
Automatically calls $(TRACE) to synchronize the H5TRACE()
macros in any source file that changed. As with makefile
dependencies, one way to force synchronization of all files is
to remove the `.depend' file.
./MANIFEST
Added new files.
./src/H5Eprivate.h
Modified HRETURN_ERROR() and HRETURN() for tracing.
./src/H5.c
./src/H5private.h
This is where the real tracing work really happens, in
H5_trace().
./src/H5A.c
./src/H5D.c
./src/H5G.c
./src/H5P.c
./src/H5S.c
./src/H5Z.c
Added H5TRACE() calls to all API functions. You don't really
need these changes if you don't want to merge your stuff
because they can be generated automatically by going to the
hdf5/src directory and saying ../bin/trace *.c
./src/H5T.c
Added H5TRACE() calls. Other stuff below.
./src/H5E.c
./src/H5Epublic.h
Added H5TRACE() calls. Created a type H5E_auto_t for the
`func' argument of H5Eset_auto() and H5Eget_auto() to make
those arguments easier to parse for tracing. It should also
make it clearer for users that don't know how to read
complicated ANSI data types.
./src/H5F.c
Added H5TRACE() calls. Changed a couple `uintn' argument
types in API functions to `unsigned int' since `uintn' part of
the API. Changed a few "can't" and "cant" error messages to
"unable to".
./src/H5Ipublic.h
Removed H5_DIRECTORY from the H5I_group_t enum. It wasn't
used anywhere.
./src/H5Tconv.c
Removed an unused label.
./src/H5Fistore.c
./src/H5Oattr.c
./src/H5Odtype.c
./src/H5T.c
./test/dsets.c
./test/dtypes.c
Fixed a warning about a variable possibly used before it's
initialized. Added __unused__ to turn off some unused
argument warnings that pop up when debugging is turned off and
optimizations are turned on.
1998-06-18 04:46:29 +08:00
dnl ----------------------------------------------------------------------
dnl Enable tracing of the API
2002-08-20 04:28:05 +08:00
dnl This must come after the enable-debug since it depends on debug.
[svn-r429] Changes since 19980616
----------------------
./html/tracing.html NEW
This entire update is to make it possible for the library to
print the name, arguments, and return value of every API call
without requiring any extra work from developers or app
programmers. This file describes how this all works.
./configure.in
Added the `--enable-tracing' switch. If you use it then the
library will include code to print API function names,
argument names and values, and function return values.
However, you must then turn on the tracing by setting the
HDF5_TRACE environment variable to a file descriptor number.
The default is `--disable-tracing' since enabling it causes a
slight increase in library size and a slowdown resulting from
an extra function call for each API function call (I couldn't
even measure the slowdown :-)
./bin/trace NEW
A perl script that synchronizes the H5TRACE() macro calls in
the *.c files with the function return type and formal
argument names and types. If you use GNU make and gcc then
this will be done automatically, otherwise just invoke this
script with the names of one or more .c files. You could do
it by hand to, but encoding argument types is a little tricky
at first.
./config/commence.in
Added the $(TRACE) macro, which defaults to the no-op.
Added -D_POSIX_SOURCE to the compiler command line.
./src/Makefile.in
Override the default for $(TRACE).
./config/depend.in
Automatically calls $(TRACE) to synchronize the H5TRACE()
macros in any source file that changed. As with makefile
dependencies, one way to force synchronization of all files is
to remove the `.depend' file.
./MANIFEST
Added new files.
./src/H5Eprivate.h
Modified HRETURN_ERROR() and HRETURN() for tracing.
./src/H5.c
./src/H5private.h
This is where the real tracing work really happens, in
H5_trace().
./src/H5A.c
./src/H5D.c
./src/H5G.c
./src/H5P.c
./src/H5S.c
./src/H5Z.c
Added H5TRACE() calls to all API functions. You don't really
need these changes if you don't want to merge your stuff
because they can be generated automatically by going to the
hdf5/src directory and saying ../bin/trace *.c
./src/H5T.c
Added H5TRACE() calls. Other stuff below.
./src/H5E.c
./src/H5Epublic.h
Added H5TRACE() calls. Created a type H5E_auto_t for the
`func' argument of H5Eset_auto() and H5Eget_auto() to make
those arguments easier to parse for tracing. It should also
make it clearer for users that don't know how to read
complicated ANSI data types.
./src/H5F.c
Added H5TRACE() calls. Changed a couple `uintn' argument
types in API functions to `unsigned int' since `uintn' part of
the API. Changed a few "can't" and "cant" error messages to
"unable to".
./src/H5Ipublic.h
Removed H5_DIRECTORY from the H5I_group_t enum. It wasn't
used anywhere.
./src/H5Tconv.c
Removed an unused label.
./src/H5Fistore.c
./src/H5Oattr.c
./src/H5Odtype.c
./src/H5T.c
./test/dsets.c
./test/dtypes.c
Fixed a warning about a variable possibly used before it's
initialized. Added __unused__ to turn off some unused
argument warnings that pop up when debugging is turned off and
optimizations are turned on.
1998-06-18 04:46:29 +08:00
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([TRACE_API])
AC_MSG_CHECKING([for API tracing]);
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([trace],
2002-08-20 04:28:05 +08:00
[AC_HELP_STRING([--enable-trace],
2003-07-29 05:38:04 +08:00
[Enable API tracing capability. Default=no
if debug is disabled.])],
2002-08-20 04:28:05 +08:00
TRACE=$enableval)
dnl Default to no if debug is disabled
if test "X-$TRACE" = X- ; then
if test -z "$DEBUG_PKG" ; then
TRACE=no
else
TRACE=yes
fi
1998-07-09 02:41:04 +08:00
fi
2003-07-29 05:38:04 +08:00
2002-08-20 04:28:05 +08:00
case "X-$TRACE" in
X-yes)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes])
2002-08-20 04:28:05 +08:00
TRACE_API=yes
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -DH5_DEBUG_API"
2002-08-20 04:28:05 +08:00
;;
X-no|*)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([no])
2002-08-20 04:28:05 +08:00
TRACE_API=no
2006-04-26 05:19:14 +08:00
H5_CPPFLAGS="$H5_CPPFLAGS -UH5_DEBUG_API"
2002-08-20 04:28:05 +08:00
;;
esac
[svn-r429] Changes since 19980616
----------------------
./html/tracing.html NEW
This entire update is to make it possible for the library to
print the name, arguments, and return value of every API call
without requiring any extra work from developers or app
programmers. This file describes how this all works.
./configure.in
Added the `--enable-tracing' switch. If you use it then the
library will include code to print API function names,
argument names and values, and function return values.
However, you must then turn on the tracing by setting the
HDF5_TRACE environment variable to a file descriptor number.
The default is `--disable-tracing' since enabling it causes a
slight increase in library size and a slowdown resulting from
an extra function call for each API function call (I couldn't
even measure the slowdown :-)
./bin/trace NEW
A perl script that synchronizes the H5TRACE() macro calls in
the *.c files with the function return type and formal
argument names and types. If you use GNU make and gcc then
this will be done automatically, otherwise just invoke this
script with the names of one or more .c files. You could do
it by hand to, but encoding argument types is a little tricky
at first.
./config/commence.in
Added the $(TRACE) macro, which defaults to the no-op.
Added -D_POSIX_SOURCE to the compiler command line.
./src/Makefile.in
Override the default for $(TRACE).
./config/depend.in
Automatically calls $(TRACE) to synchronize the H5TRACE()
macros in any source file that changed. As with makefile
dependencies, one way to force synchronization of all files is
to remove the `.depend' file.
./MANIFEST
Added new files.
./src/H5Eprivate.h
Modified HRETURN_ERROR() and HRETURN() for tracing.
./src/H5.c
./src/H5private.h
This is where the real tracing work really happens, in
H5_trace().
./src/H5A.c
./src/H5D.c
./src/H5G.c
./src/H5P.c
./src/H5S.c
./src/H5Z.c
Added H5TRACE() calls to all API functions. You don't really
need these changes if you don't want to merge your stuff
because they can be generated automatically by going to the
hdf5/src directory and saying ../bin/trace *.c
./src/H5T.c
Added H5TRACE() calls. Other stuff below.
./src/H5E.c
./src/H5Epublic.h
Added H5TRACE() calls. Created a type H5E_auto_t for the
`func' argument of H5Eset_auto() and H5Eget_auto() to make
those arguments easier to parse for tracing. It should also
make it clearer for users that don't know how to read
complicated ANSI data types.
./src/H5F.c
Added H5TRACE() calls. Changed a couple `uintn' argument
types in API functions to `unsigned int' since `uintn' part of
the API. Changed a few "can't" and "cant" error messages to
"unable to".
./src/H5Ipublic.h
Removed H5_DIRECTORY from the H5I_group_t enum. It wasn't
used anywhere.
./src/H5Tconv.c
Removed an unused label.
./src/H5Fistore.c
./src/H5Oattr.c
./src/H5Odtype.c
./src/H5T.c
./test/dsets.c
./test/dtypes.c
Fixed a warning about a variable possibly used before it's
initialized. Added __unused__ to turn off some unused
argument warnings that pop up when debugging is turned off and
optimizations are turned on.
1998-06-18 04:46:29 +08:00
2004-07-23 04:46:27 +08:00
dnl ----------------------------------------------------------------------
dnl Enable instrumenting of the library's internal operations
dnl This must come after the enable-debug since it depends on debug.
dnl
AC_SUBST([INSTRUMENT_LIBRARY])
AC_MSG_CHECKING([for instrumented library]);
AC_ARG_ENABLE([instrument],
[AC_HELP_STRING([--enable-instrument],
[Enable library instrumentation of optimization
tracing. Default=no if debug is disabled.])],
INSTRUMENT=$enableval)
dnl Default to no if debug is disabled
if test "X-$INSTRUMENT" = X- ; then
if test -z "$DEBUG_PKG" ; then
INSTRUMENT=no
else
INSTRUMENT=yes
fi
fi
case "X-$INSTRUMENT" in
X-yes)
AC_MSG_RESULT([yes])
INSTRUMENT_LIBRARY=yes
AC_DEFINE([HAVE_INSTRUMENTED_LIBRARY], [1],
[Define if library will contain instrumentation to detect correct optimization operation])
;;
X-no|*)
AC_MSG_RESULT([no])
INSTRUMENT_LIBRARY=no
;;
esac
2007-05-30 04:02:34 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they would like to securely clear file buffers before they are
dnl written.
dnl
AC_SUBST([CLEARFILEBUF])
AC_MSG_CHECKING([whether to clear file buffers])
AC_ARG_ENABLE([clear-file-buffers],
[AC_HELP_STRING([--enable-clear-file-buffers],
[Securely clear file buffers before writing
to file. Default=yes.])],
[CLEARFILEBUF=$enableval])
case "X-$CLEARFILEBUF" in
*)
CLEARFILEBUF=yes
AC_MSG_RESULT([yes])
AC_DEFINE([CLEAR_MEMORY], [1],
[Define if the memory buffers being written to disk should be
cleared before writing.])
;;
X-no)
CLEARFILEBUF=no
AC_MSG_RESULT([no])
;;
esac
dnl ----------------------------------------------------------------------
dnl Check if they would like to use a memory checking tool (like valgrind's
dnl 'memcheck' tool, or Rational Purify, etc) and the library should be
dnl more scrupulous with it's memory operations. Enabling this also
dnl disables the library's free space manager code.
dnl
AC_SUBST([USINGMEMCHECKER])
AC_MSG_CHECKING([whether a memory checking tool will be used])
AC_ARG_ENABLE([using-memchecker],
[AC_HELP_STRING([--enable-using-memchecker],
[Enable this option if a memory allocation and/or
bounds checking tool will be used on the HDF5
library. Enabling this causes the library to be
more picky about it's memory operations and also
disables the library's free space manager code.
Default=no.])],
[USINGMEMCHECKER=$enableval])
case "X-$USINGMEMCHECKER" in
X-yes)
USINGMEMCHECKER=yes
AC_MSG_RESULT([yes])
AC_DEFINE([USING_MEMCHECKER], [1],
[Define if a memory checking tool will be used on the library,
to cause library to be very picky about memory operations and
also disable the internal free list manager code.])
;;
*)
USINGMEMCHECKER=no
AC_MSG_RESULT([no])
;;
esac
2000-04-21 02:41:51 +08:00
dnl Checkpoint the cache
AC_CACHE_SAVE
1999-03-04 07:17:48 +08:00
dnl What header files and libraries do we have to look for for parallel
dnl support? For the most part, search paths are already specified with
dnl CPPFLAGS and LDFLAGS or are known to the compiler. If the user says
dnl `--disable-parallel' but specifies a known parallel compiler (like mpicc
dnl or mpcc) then parallel support is enabled but configure doesn't search
dnl for any parallel header files or libraries.
dnl
2002-07-18 00:10:46 +08:00
AC_ARG_ENABLE([parallel],
2003-04-10 07:24:30 +08:00
[AC_HELP_STRING([--enable-parallel],
2002-07-18 00:10:46 +08:00
[Search for MPI-IO and MPI support files])])
1999-03-04 07:17:48 +08:00
2005-09-02 04:31:46 +08:00
dnl The --enable-parallel flag is not compatible with --enable-cxx.
2010-11-18 02:07:51 +08:00
dnl If the user tried to specify both flags, throw an error, unless
dnl they also provided the --enable-unsupported flag.
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
if test "X${HDF_CXX}" = "Xyes" -a "X${enable_parallel}" = "Xyes"; then
AC_MSG_ERROR([--enable-cxx and --enable-parallel flags are incompatible. Use --enable-unsupported to override this error.])
fi
2005-09-02 04:31:46 +08:00
fi
2010-11-18 02:07:51 +08:00
dnl --enable-parallel is also incompatible with --enable-threadsafe, unless
dnl --enable-unsupported has been specified on the configure line.
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
if test "X${THREADSAFE}" = "Xyes" -a "X${enable_parallel}" = "Xyes"; then
AC_MSG_ERROR([--enable-threadsafe and --enable-parallel flags are incompatible. Use --enable-unsupported to override this error.])
fi
2005-09-02 04:31:46 +08:00
fi
2005-09-07 00:57:05 +08:00
dnl It's possible to build in parallel by specifying a parallel compiler
dnl without using the --enable-parallel flag. This isn't allowed with
2010-11-18 02:07:51 +08:00
dnl C++ or threadsafe, either, unless the --enable-unsupported flag
dnl has also been specified.
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
if test "X${PARALLEL}" != "X" -a "X${enable_cxx}" = "Xyes" ; then
AC_MSG_ERROR([An MPI compiler is being used; --enable-cxx is not allowed. Use --enable-unsupported to override this error.])
fi
if test "X${PARALLEL}" != "X" -a "X${THREADSAFE}" = "Xyes"; then
AC_MSG_ERROR([An MPI compiler is being used; --enable-threadsafe is not allowed. Use --enable-unsupported to override this error.])
fi
2005-09-07 00:57:05 +08:00
fi
2002-07-18 00:10:46 +08:00
AC_MSG_CHECKING([for parallel support files])
1999-03-04 07:17:48 +08:00
case "X-$enable_parallel" in
2001-09-28 07:05:16 +08:00
X-|X-no|X-none)
dnl Either we are not compiling for parallel or the header and
dnl library files and locations are known to the compiler (this is
dnl the case for a correct installation of mpicc for instance).
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([skipped])
2001-09-28 07:05:16 +08:00
;;
2000-09-15 06:19:24 +08:00
2001-09-28 07:05:16 +08:00
X-yes)
dnl We want to compile a parallel library with a compiler that
dnl may already know how to link with MPI and MPI-IO.
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([provided by compiler])
2001-09-28 07:05:16 +08:00
PARALLEL=yes
2005-09-29 12:57:44 +08:00
dnl Try link a simple MPI program. If fail, try again with -lmpi and
dnl -lmpich.
AC_TRY_LINK(, MPI_Init(),, \
AC_CHECK_LIB(mpi, MPI_Init,, \
AC_CHECK_LIB(mpich, MPI_Init,, PARALLEL=no)))
2001-09-28 07:05:16 +08:00
dnl Then try link a simple MPI-IO program. If fail, try again with
dnl -lmpio.
if test "X$PARALLEL" = "Xyes"; then
2003-07-29 05:38:04 +08:00
AC_TRY_LINK(, [MPI_File_open()],,
[AC_CHECK_LIB([mpio], [MPI_File_open],, [PARALLEL=no])])
fi
if test "X$HDF_FORTRAN" = "Xyes"; then
dnl Change to the Fortran 90 language
2005-03-11 04:52:34 +08:00
AC_LANG_PUSH(Fortran)
2003-07-29 05:38:04 +08:00
dnl Try link a simple MPI program. If fail, try again with -lmpi.
2006-03-31 07:08:50 +08:00
AC_LINK_IFELSE([
2006-04-01 02:51:49 +08:00
program main
2006-03-29 06:13:46 +08:00
include 'mpif.h'
2004-10-13 04:49:28 +08:00
integer:: ierr
2006-04-01 02:51:49 +08:00
call mpi_file_open( ierr )
end],,
2006-03-29 06:13:46 +08:00
AC_CHECK_LIB(mpi, [
2003-07-29 05:38:04 +08:00
include 'mpif.h'
2004-10-13 04:49:28 +08:00
integer:: ierr
2003-07-29 05:38:04 +08:00
call mpi_file_open( ierr )],, PARALLEL=no))
dnl Then try link a simple MPI-IO program. If fail, try again with
dnl -lmpio.
if test "X$PARALLEL" = "Xyes"; then
2006-03-31 07:08:50 +08:00
AC_LINK_IFELSE([
2006-04-01 02:51:49 +08:00
program main
2006-03-29 06:13:46 +08:00
include 'mpif.h'
2004-10-13 04:49:28 +08:00
integer:: ierr
2006-04-01 02:51:49 +08:00
call mpi_file_open( ierr )
end],,
2006-03-29 06:13:46 +08:00
AC_CHECK_LIB(mpio, [
2003-07-29 05:38:04 +08:00
include 'mpif.h'
2004-10-13 04:49:28 +08:00
integer:: ierr
2003-07-29 05:38:04 +08:00
call mpi_file_open( ierr )],, PARALLEL=no))
fi
dnl Change to the C language
2005-03-11 04:52:34 +08:00
AC_LANG_POP(Fortran)
2001-09-28 07:05:16 +08:00
fi
2010-06-12 07:37:33 +08:00
dnl Set RUNPARALLEL to mpiexec if not set yet.
2009-04-30 10:40:32 +08:00
dnl Check for building on Cray if RUNPARALLEL is not yet set by checking
2010-06-12 07:37:33 +08:00
dnl for 'aprun' command (which is the parallel job launcher, like mpiexec).
2009-04-30 10:40:32 +08:00
if test "X$PARALLEL" = "Xyes" -a -z "$RUNPARALLEL"; then
dnl Find the path where aprun is located.
for path in `echo $PATH | ${TR} ":" " "`; do
if test -x $path/aprun; then
2010-10-29 01:19:30 +08:00
RUNPARALLEL="aprun -q -n \$\${NPROCS:=6}"
2009-04-30 10:40:32 +08:00
break;
fi
done
fi
2010-06-12 07:37:33 +08:00
dnl Set RUNPARALLEL to mpiexec if not set yet.
2001-09-28 07:05:16 +08:00
if test "X$PARALLEL" = "Xyes" -a -z "$RUNPARALLEL"; then
2010-10-29 01:19:30 +08:00
RUNPARALLEL="mpiexec -n \$\${NPROCS:=6}"
2001-09-28 07:05:16 +08:00
fi
;;
*)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([error])
AC_MSG_ERROR([\'$enable_parallel\' is not a valid parallel search type])
2001-09-28 07:05:16 +08:00
;;
1999-03-04 07:17:48 +08:00
esac
dnl ----------------------------------------------------------------------
2003-07-29 05:38:04 +08:00
dnl Print some other parallel information and do some sanity checks.
1999-03-04 07:17:48 +08:00
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([ADD_PARALLEL_FILES]) ADD_PARALLEL_FILES="no"
2000-09-15 06:19:24 +08:00
if test -n "$PARALLEL"; then
2003-08-09 02:28:03 +08:00
dnl The 'testpar' directory should participate in the build
2001-09-28 07:05:16 +08:00
TESTPARALLEL=testpar
1999-03-04 07:17:48 +08:00
2001-09-28 07:05:16 +08:00
dnl We are building a parallel library
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_PARALLEL], [1], [Define if we have parallel support])
2001-09-28 07:05:16 +08:00
dnl Display what we found about running programs
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([prefix for running on one processor])
AC_MSG_RESULT([$RUNSERIAL])
AC_MSG_CHECKING([prefix for running in parallel])
AC_MSG_RESULT([$RUNPARALLEL])
2001-09-28 07:05:16 +08:00
dnl Check that we can link a simple MPI and MPI-IO application
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([whether a simple MPI-IO program can be linked])
AC_TRY_LINK(, [MPI_Init(); MPI_File_open();],
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no])
AC_MSG_ERROR([unable to link a simple MPI-IO application]))
2001-09-28 07:05:16 +08:00
dnl There *must* be some way to run in parallel even if it's just the
dnl word `none'.
if test -z "$RUNPARALLEL"; then
2003-07-29 05:38:04 +08:00
AC_MSG_ERROR([no way to run a parallel program])
2001-09-28 07:05:16 +08:00
fi
dnl If RUNSERIAL or RUNPARALLEL is the word `none' then replace it with
dnl the empty string.
if test "X$RUNSERIAL" = "Xnone"; then
RUNSERIAL=""
fi
if test "X$RUNPARALLEL" = "Xnone"; then
RUNPARALLEL=""
fi
2003-07-29 05:38:04 +08:00
if test "X$HDF_FORTRAN" = "Xyes"; then
ADD_PARALLEL_FILES="yes"
AC_MSG_CHECKING([for MPI_Comm_c2f and MPI_Comm_f2c functions])
AC_TRY_LINK([#include <mpi.h>],
[MPI_Comm c_comm; MPI_Comm_c2f(c_comm)],
AC_DEFINE([HAVE_MPI_MULTI_LANG_Comm], [1],
2005-02-01 11:17:02 +08:00
[Define if `MPI_Comm_c2f' and `MPI_Comm_f2c' exists])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no])
)
AC_MSG_CHECKING([for MPI_Info_c2f and MPI_Info_f2c functions])
AC_TRY_LINK([#include <mpi.h>],
[MPI_Info c_info; MPI_Info_c2f(c_info)],
AC_DEFINE([HAVE_MPI_MULTI_LANG_Info], [1],
2005-02-01 11:17:02 +08:00
[Define if `MPI_Info_c2f' and `MPI_Info_f2c' exists])
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([yes]),
AC_MSG_RESULT([no])
)
fi
2003-06-24 23:34:27 +08:00
dnl --------------------------------------------------------------------
dnl Do we want MPE instrumentation feature on?
dnl
dnl This must be done after enable-parallel is checked since it depends
dnl on a mpich compiler.
dnl
2008-05-15 13:12:00 +08:00
MPE=yes
2003-06-24 23:34:27 +08:00
AC_ARG_WITH([mpe],
[AC_HELP_STRING([--with-mpe=DIR],
[Use MPE instrumentation [default=no]])],,
[withval=no])
case "X-$withval" in
X-|X-no|X-none)
AC_MSG_CHECKING([for MPE])
AC_MSG_RESULT([suppressed])
unset MPE
;;
X-yes)
AC_CHECK_HEADERS([mpe.h],, [unset MPE])
AC_CHECK_LIB([mpe], [MPE_Init_log],, [unset MPE])
2003-06-25 03:24:01 +08:00
AC_CHECK_LIB([lmpe], [CLOG_Init],, [unset MPE])
2003-06-24 23:34:27 +08:00
;;
*)
case "$withval" in
*,*)
mpe_inc="`echo $withval | cut -f1 -d,`"
mpe_lib="`echo $withval | cut -f2 -d, -s`"
;;
*)
if test -n "$withval"; then
mpe_inc="$withval/include"
mpe_lib="$withval/lib"
fi
;;
esac
dnl Trying to include -I/usr/include and -L/usr/lib is redundant and
dnl can mess some compilers up.
if test "X$mpe_inc" = "X/usr/include"; then
mpe_inc=""
fi
if test "X$mpe_lib" = "X/usr/lib"; then
mpe_lib=""
fi
if test -n "$mpe_inc"; then
saved_CPPFLAGS="$CPPFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_CPPFLAGS="$AM_CPPFLAGS"
2003-06-24 23:34:27 +08:00
CPPFLAGS="$CPPFLAGS -I$mpe_inc"
2009-10-15 05:43:05 +08:00
AM_CPPFLAGS="$AM_CPPFLAGS -I$mpe_inc"
AC_CHECK_HEADERS([mpe.h],, [CPPFLAGS="$saved_CPPFLAGS"; AM_CPPFLAGS="$saved_AM_CPPFLAGS"; unset MPE])
2003-06-24 23:34:27 +08:00
else
AC_CHECK_HEADERS([mpe.h],, [unset MPE])
fi
if test -n "$mpe_lib"; then
saved_LDFLAGS="$LDFLAGS"
2009-10-15 05:43:05 +08:00
saved_AM_LDFLAGS="$AM_LDFLAGS"
2003-06-24 23:34:27 +08:00
LDFLAGS="$LDFLAGS -L$mpe_lib"
2009-10-15 05:43:05 +08:00
AM_LDFLAGS="$AM_LDFLAGS -L$mpe_lib"
2003-06-24 23:34:27 +08:00
AC_CHECK_LIB([mpe], [MPE_Init_log],,
2009-10-15 05:43:05 +08:00
[LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset MPE])
2003-06-25 03:24:01 +08:00
AC_CHECK_LIB([lmpe], [CLOG_Init],,
2009-10-15 05:43:05 +08:00
[LDFLAGS="$saved_LDFLAGS"; AM_LDFLAGS="$saved_AM_LDFLAGS"; unset MPE])
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2003-06-24 23:34:27 +08:00
else
AC_CHECK_LIB([mpe], [MPE_Init_log],, [unset MPE])
2003-06-25 03:24:01 +08:00
AC_CHECK_LIB([lmpe], [CLOG_Init],, [unset MPE])
2003-06-24 23:34:27 +08:00
fi
;;
esac
if test "X-$MPE" = "X-yes"; then
AC_DEFINE([HAVE_MPE], [1], [Define if we have MPE support])
fi
2003-06-17 01:53:12 +08:00
2003-09-17 01:33:00 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the MPI_File_set_size() function
dnl works with files over 2GB, unless it's already set in the cache.
dnl (This flag should be set for all machines, except for ASCI Red, where
dnl the cache value is set in it's config file)
dnl
AC_MSG_CHECKING([if MPI_File_set_size works for files over 2GB])
AC_CACHE_VAL([hdf5_cv_mpi_file_set_size_big], [hdf5_cv_mpi_file_set_size_big=yes])
if test ${hdf5_cv_mpi_file_set_size_big} = "yes"; then
AC_DEFINE([MPI_FILE_SET_SIZE_BIG], [1],
2005-02-01 11:17:02 +08:00
[Define if your system's `MPI_File_set_size' function works for files over 2GB.])
2003-09-17 01:33:00 +08:00
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-03-12 06:11:05 +08:00
2007-06-03 02:19:36 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the MPI_File_get_size() function
dnl works. The default is enabled unless the user knows the function
dnl doesn't work on the system and disables it. (This flag should be set
dnl for all machines except for SGI Altix Propack 4 where the function
dnl doesn't return correct file size.)
dnl
AC_ARG_ENABLE([mpi-size],
[AC_HELP_STRING([--enable-mpi-size],
[Some systems (only SGI Altix Propack 4 so far) return wrong value
from MPI_File_get_size. By disabling this function, the library
will replace it with stat to get the correct file size.
[default=yes]])],
[MPI_GET_SIZE=$enableval])
AC_MSG_CHECKING([if MPI_File_get_size is enabled])
AC_SUBST(MPI_GET_SIZE)
case "X-$MPI_GET_SIZE" in
X-no)
AC_MSG_RESULT([no])
MPI_GET_SIZE=no
;;
X-yes|*)
AC_MSG_RESULT([yes])
MPI_GET_SIZE=yes
AC_DEFINE([HAVE_MPI_GET_SIZE], [1],
[Define if MPI_File_get_size works correctly])
;;
esac
2005-03-12 06:11:05 +08:00
dnl ----------------------------------------------------------------------
dnl Check to see whether the complicate MPI derived datatype works.
2006-02-17 00:45:42 +08:00
dnl In Dec. 20th, 2004, we found that IBM's MPIO implemention didn't
2005-03-12 06:11:05 +08:00
dnl handle with the displacement of the complicate MPI type derived datatype
dnl correctly. So we add the check here.
2006-02-17 00:45:42 +08:00
dnl IBM fixed this bug in their new version that supported MPI-IO around spring 2005.
dnl We find that mpich 1.2.5 has the similar bug. The same
dnl bug also occurs at SGI IRIX 6.5 C with compiler version lower than or equal to 7.3.
dnl In case people still use the old compiler, we keep this flag.
2005-03-12 06:11:05 +08:00
AC_MSG_CHECKING([if irregular hyperslab optimization code works inside MPI-IO])
2009-10-16 05:14:27 +08:00
AC_CACHE_VAL([hdf5_cv_mpi_complex_derived_datatype_works],[hdf5_cv_mpi_complex_derived_datatype_works=yes])
2005-03-12 06:11:05 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_mpi_complex_derived_datatype_works} = "yes"; then
2005-03-12 06:11:05 +08:00
AC_DEFINE([MPI_COMPLEX_DERIVED_DATATYPE_WORKS], [1],
[Define if your system can handle complicated MPI derived datatype correctly.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2006-02-17 00:45:42 +08:00
2006-04-27 05:19:21 +08:00
2006-02-17 00:45:42 +08:00
dnl ----------------------------------------------------------------------
dnl Check to see whether MPI-IO can do collective IO successfully when one or more processes don't do
dnl any IOs.
dnl Up to now(Feb. 8th, 2006), we find that it will cause program hung with mpich 1.2.x version
dnl and SGI altix. For those systems, we have to turn off this feature and use independent IO instead.
dnl
AC_MSG_CHECKING([if MPI-IO can do collective IO when one or more processes don't do IOs])
2009-10-16 05:14:27 +08:00
AC_CACHE_VAL([hdf5_cv_mpi_special_collective_io_works],[hdf5_cv_mpi_special_collective_io_works=yes])
2006-02-17 00:45:42 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_mpi_special_collective_io_works} = "yes"; then
2006-02-17 00:45:42 +08:00
AC_DEFINE([MPI_SPECIAL_COLLECTIVE_IO_WORKS], [1],
[Define if your system can handle special collective IO properly.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2003-09-17 01:33:00 +08:00
fi
2004-06-09 21:47:20 +08:00
2002-11-12 04:32:02 +08:00
dnl ----------------------------------------------------------------------
dnl Turn on internal I/O filters by setting macros in header files
dnl Internal I/O filters are contained entirely within the library and do
dnl not depend on external headers or libraries. The shuffle filter is
dnl an example of an internal filter, while the gzip filter is an example of
dnl an external filter. Each external filter is controlled with an
dnl "--with-foo=" configure flag.
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([FILTERS])
2004-06-09 21:47:20 +08:00
AC_SUBST(USE_FILTER_SHUFFLE) USE_FILTER_SHUFFLE="no"
AC_SUBST(USE_FILTER_FLETCHER32) USE_FILTER_FLETCHER32="no"
2005-01-07 02:48:49 +08:00
AC_SUBST(USE_FILTER_NBIT) USE_FILTER_NBIT="no"
2005-02-17 09:50:42 +08:00
AC_SUBST(USE_FILTER_SCALEOFFSET) USE_FILTER_SCALEOFFSET="no"
2003-07-29 05:38:04 +08:00
AC_MSG_CHECKING([for I/O filters])
2002-11-12 04:32:02 +08:00
AC_ARG_ENABLE([filters],
[AC_HELP_STRING([--enable-filters=all],
[Turn on all internal I/O filters. One may
also specify a comma-separated list of filters
or the word no. The default is all internal
I/O filters.])],
[FILTERS=$enableval])
dnl Eventually: all_filters="shuffle,foo,bar,baz"
2005-02-17 09:50:42 +08:00
all_filters="shuffle,fletcher32,nbit,scaleoffset"
2002-11-12 04:32:02 +08:00
case "X-$FILTERS" in
X-|X-all)
FILTERS=$all_filters
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([all ($FILTERS)])
2002-11-12 04:32:02 +08:00
;;
X-no|X-none)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([none])
2002-11-12 04:32:02 +08:00
FILTERS="none"
;;
*)
2003-07-29 05:38:04 +08:00
AC_MSG_RESULT([$FILTERS])
2002-11-12 04:32:02 +08:00
;;
esac
if test -n "$FILTERS"; then
2006-05-19 20:22:07 +08:00
for filter in `echo $FILTERS | tr ${as_cr_letters}',' ${as_cr_LETTERS}' '`; do
2003-07-29 05:38:04 +08:00
dnl ------------------------------------------------------------------
dnl Have to use separate 'if' construct for each filter, so that
dnl autoheader can detect the AC_DEFINE for each one...
dnl
2002-11-12 04:32:02 +08:00
if test $filter = "SHUFFLE"; then
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_FILTER_SHUFFLE], [1],
2002-11-12 04:32:02 +08:00
[Define if support for shuffle filter is enabled])
2004-06-09 21:47:20 +08:00
USE_FILTER_SHUFFLE="yes"
2002-11-12 04:32:02 +08:00
fi
2003-02-13 04:07:21 +08:00
if test $filter = "FLETCHER32"; then
2003-07-29 05:38:04 +08:00
AC_DEFINE([HAVE_FILTER_FLETCHER32], [1],
2003-02-13 04:07:21 +08:00
[Define if support for Fletcher32 checksum is enabled])
2004-06-09 21:47:20 +08:00
USE_FILTER_FLETCHER32="yes"
2003-02-05 02:50:56 +08:00
fi
2005-01-07 02:48:49 +08:00
if test $filter = "NBIT"; then
AC_DEFINE([HAVE_FILTER_NBIT], [1],
[Define if support for nbit filter is enabled])
USE_FILTER_NBIT="yes"
fi
2005-02-17 09:50:42 +08:00
if test $filter = "SCALEOFFSET"; then
AC_DEFINE([HAVE_FILTER_SCALEOFFSET], [1],
[Define if support for scaleoffset filter is enabled])
USE_FILTER_SCALEOFFSET="yes"
fi
2002-11-12 04:32:02 +08:00
done
fi
2003-05-06 05:13:49 +08:00
dnl ----------------------------------------------------------------------
dnl This is defined only when we're using CodeWarrior, since it has a
dnl broken "open()" call.
dnl
if test 1 = 2; then
2003-07-29 05:38:04 +08:00
AC_DEFINE([NO_SHARED_WRITING], [1],
2003-05-06 05:13:49 +08:00
[Define if shared writing must be disabled (CodeWarrior only)])
fi
2007-06-11 21:35:43 +08:00
dnl --------------------------------------------------------------------------
dnl Should the Default Virtual File Driver be compiled?
dnl
AC_MSG_CHECKING([for Default Virtual File Driver definition])
AC_ARG_WITH([default-vfd],
[AC_HELP_STRING([--with-default-vfd=driver],
[Specify default file driver
2007-06-12 02:51:20 +08:00
[default=sec2]])],,
withval=sec2)
2007-06-11 21:35:43 +08:00
if test "X$withval" = "Xsec2"; then
AC_MSG_RESULT([yes])
default_vfd=yes
vfd_define=H5FD_SEC2
elif test "X$withval" = "Xstdio"; then
AC_MSG_RESULT([yes])
default_vfd=yes
vfd_define=H5FD_STDIO
else
AC_MSG_RESULT([no])
default_vfd=no
fi
if test "X$default_vfd" = "Xyes"; then
AC_DEFINE_UNQUOTED([DEFAULT_VFD], [$vfd_define],
[Define the default virtual file driver to compile])
fi
2003-05-06 05:13:49 +08:00
2006-10-11 04:07:16 +08:00
dnl ----------------------------------------------------------------------
2006-11-07 06:17:46 +08:00
dnl Check if Direct I/O driver is enabled by --enable-direct-vfd
2006-10-11 04:07:16 +08:00
dnl
AC_MSG_CHECKING([for Direct Virtual File Driver support])
AC_ARG_ENABLE([direct-vfd],
[AC_HELP_STRING([--enable-direct-vfd],
[Build the Direct I/O Virtual File Driver
[default=yes]])],
[DIRECT_VFD=$enableval], [DIRECT_VFD=yes])
if test "$DIRECT_VFD" = "yes"; then
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_direct_io],
2006-10-11 04:07:16 +08:00
[AC_TRY_RUN([
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
int main(void)
{
int fid;
if((fid=open("tst_file", O_CREAT | O_TRUNC | O_DIRECT, 0755))<0)
exit(1);
close(fid);
remove("tst_file");
exit (0);
2009-10-16 05:14:27 +08:00
}], AC_TRY_LINK(, [posix_memalign()], [hdf5_cv_direct_io=yes], [hdf5_cv_direct_io=no]), [hdf5_cv_direct_io=no],)])
2006-10-11 04:07:16 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_direct_io} = "yes"; then
2006-10-11 04:07:16 +08:00
AC_MSG_RESULT([yes])
AC_DEFINE([HAVE_DIRECT], [1],
[Define if the direct I/O virtual file driver should be compiled])
else
AC_MSG_RESULT([no])
2006-10-11 04:47:24 +08:00
DIRECT_VFD=no
2006-10-11 04:07:16 +08:00
fi
else
AC_MSG_RESULT([suppressed])
fi
2006-10-11 04:47:24 +08:00
AM_CONDITIONAL([DIRECT_VFD_CONDITIONAL], [test "X$DIRECT_VFD" = "Xyes"])
2006-11-07 06:17:46 +08:00
2005-08-29 23:21:58 +08:00
dnl ----------------------------------------------------------------------
dnl Decide whether the presence of user's exception handling functions is
dnl checked and data conversion exceptions are returned. This is mainly
dnl for the speed optimization of hard conversions. Soft conversions can
dnl actually benefit little.
dnl
AC_MSG_CHECKING([whether exception handling functions is checked during data conversions])
2005-08-30 23:11:08 +08:00
AC_ARG_ENABLE([dconv-exception],
[AC_HELP_STRING([--enable-dconv-exception],
2005-08-29 23:21:58 +08:00
[if exception handling functions is checked during
data conversions [default=yes]])],
2005-09-02 02:48:00 +08:00
[DCONV_EXCEPTION=$enableval], [DCONV_EXCEPTION=yes])
2005-08-29 23:21:58 +08:00
if test "$DCONV_EXCEPTION" = "yes"; then
AC_MSG_RESULT([yes])
AC_DEFINE([WANT_DCONV_EXCEPTION], [1],
[Check exception handling functions during data conversions])
else
AC_MSG_RESULT([no])
fi
dnl ----------------------------------------------------------------------
dnl Decide whether the data accuracy has higher priority during data
dnl conversions. If not, some hard conversions will still be prefered even
dnl though the data may be wrong (for example, some compilers don't
dnl support denormalized floating values) to maximize speed.
dnl
AC_MSG_CHECKING([whether data accuracy is guaranteed during data conversions])
2005-08-30 23:11:08 +08:00
AC_ARG_ENABLE([dconv-accuracy],
[AC_HELP_STRING([--enable-dconv-accuracy],
2005-08-29 23:21:58 +08:00
[if data accuracy is guaranteed during
data conversions [default=yes]])],
2005-09-02 02:48:00 +08:00
[DATA_ACCURACY=$enableval], [DATA_ACCURACY=yes])
2005-08-29 23:21:58 +08:00
if test "$DATA_ACCURACY" = "yes"; then
AC_MSG_RESULT([yes])
AC_DEFINE([WANT_DATA_ACCURACY], [1],
[Data accuracy is prefered to speed during data conversions])
else
AC_MSG_RESULT([no])
fi
2004-01-01 04:36:37 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can handle converting
dnl denormalized floating-point values.
dnl (This flag should be set for all machines, except for the Crays, where
dnl the cache value is set in it's config file)
dnl
AC_MSG_CHECKING([if converting denormalized floating-point values is possible])
AC_CACHE_VAL([hdf5_cv_convert_denormal_float], [hdf5_cv_convert_denormal_float=yes])
if test ${hdf5_cv_convert_denormal_float} = "yes"; then
AC_DEFINE([CONVERT_DENORMAL_FLOAT], [1],
[Define if your system can handle converting denormalized floating-point values.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-11-15 10:55:39 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can handle converting
dnl floating-point to long long values.
dnl (This flag should be _unset_ for all machines, except for Windows, where
dnl it's set in the custom Windows H5pubconf.h file)
dnl
AC_MSG_CHECKING([if converting floating-point values to long long is not working])
2006-06-02 02:34:32 +08:00
AC_CACHE_VAL([hdf5_cv_convert_float_llong_not_works], [hdf5_cv_convert_float_llong_not_works=no])
2005-11-15 10:55:39 +08:00
if test ${hdf5_cv_convert_float_llong_not_works} = "yes"; then
AC_DEFINE([HW_FP_TO_LLONG_NOT_WORKS], [1],
[Define if your system can't handle converting floating-point values to long long.])
2006-06-02 02:34:32 +08:00
AC_MSG_RESULT([true])
2005-11-15 10:55:39 +08:00
else
2006-06-02 02:34:32 +08:00
AC_MSG_RESULT([false])
2005-11-15 10:55:39 +08:00
fi
2008-05-01 03:51:13 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine has window style pathname,
dnl that is, "drive-letter:\" (e.g. "C:") or "drive-letter:/" (e.g. "C:/").
dnl (This flag should be _unset_ for all machines, except for Windows, where
dnl it's set in the custom Windows H5pubconf.h file)
dnl
AC_MSG_CHECKING([if the machine has window style path name])
2008-05-16 05:35:13 +08:00
case "`uname`" in
MINGW*)
AC_DEFINE([HAVE_WINDOW_PATH], [1],
[Define if your system has window style path name.])
AC_MSG_RESULT([yes])
;;
*)
AC_MSG_RESULT([no])
;;
esac
2008-05-01 03:51:13 +08:00
2005-01-07 02:17:59 +08:00
dnl -----------------------------------------------------------------------
2005-01-09 04:56:12 +08:00
dnl Set flag to indicate that the machine can handle conversion from
2005-08-29 23:21:58 +08:00
dnl long double to integers accurately. This flag should be set "yes" for
dnl all machines except all SGIs. For SGIs, some conversions are
2005-01-09 04:56:12 +08:00
dnl incorrect and its cache value is set "no" in its config/irix6.x and
2005-08-29 23:21:58 +08:00
dnl irix5.x.
dnl
AC_MSG_CHECKING([if converting from long double to integers is accurate])
2005-01-12 05:05:21 +08:00
2005-08-29 23:21:58 +08:00
if test ${ac_cv_sizeof_long_double} = 0; then
2009-10-16 05:14:27 +08:00
hdf5_cv_ldouble_to_integer_accurate=${hdf5_cv_ldouble_to_integer_accurate=no}
2005-08-29 23:21:58 +08:00
else
AC_CACHE_VAL([hdf5_cv_ldouble_to_integer_accurate], [hdf5_cv_ldouble_to_integer_accurate=yes])
fi
2005-01-12 05:05:21 +08:00
2005-09-21 23:07:44 +08:00
if test "${hdf5_cv_ldouble_to_integer_accurate}" = "yes"; then
2005-08-29 23:21:58 +08:00
AC_DEFINE([LDOUBLE_TO_INTEGER_ACCURATE], [1],
[Define if your system can convert long double to integers accurately.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
2005-07-23 09:55:12 +08:00
fi
2005-08-29 23:21:58 +08:00
dnl -----------------------------------------------------------------------
dnl Set flag to indicate that the machine can do conversion from
dnl long double to integers regardless of accuracy. This flag should be
dnl set "yes" for all machines except HP-UX 11.00. For HP-UX 11.00, the
dnl compiler has 'floating exception' when converting 'long double' to all
dnl integers except 'unsigned long long'. Other HP-UX systems are unknown
dnl yet. (1/8/05 - SLU)
2005-01-12 05:05:21 +08:00
2005-08-29 23:21:58 +08:00
AC_MSG_CHECKING([if converting from long double to integers works])
2005-01-07 02:17:59 +08:00
2005-08-29 23:21:58 +08:00
if test ${ac_cv_sizeof_long_double} = 0; then
2009-10-16 05:14:27 +08:00
hdf5_cv_ldouble_to_integer_works=${hdf5_cv_ldouble_to_integer_works=no}
2005-08-29 23:21:58 +08:00
else
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_ldouble_to_integer_works],
2005-08-29 23:21:58 +08:00
[AC_TRY_RUN([
int main(void)
{
void *align;
long double ld= 9701917572145405952.00L;
unsigned char v1;
short v2;
unsigned int v3;
int ret = 0;
align = (void*)malloc(sizeof(long double));
memcpy(align, &ld, sizeof(long double));
/*For HU-UX11.00, there's floating exception(core dump) when doing some of casting
*from 'long double' to integers*/
v1=(unsigned char)(*((long double*)align));
v2=(short)(*((long double*)align));
v3=(unsigned int)(*((long double*)align));
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_ldouble_to_integer_works=yes], [hdf5_cv_ldouble_to_integer_works=no],)])
2005-08-29 23:21:58 +08:00
fi
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_ldouble_to_integer_works} = "yes"; then
2005-08-29 23:21:58 +08:00
AC_DEFINE([LDOUBLE_TO_INTEGER_WORKS], [1],
[Define if your system can convert from long double to integer values.])
2005-01-09 04:56:12 +08:00
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
dnl -----------------------------------------------------------------------
dnl Set flag to indicate that the machine can handle conversion from
dnl integers to long double. (This flag should be set "yes" for all
dnl machines except all SGIs, where some conversions are
dnl incorrect and its cache value is set "no" in its config/irix6.x and
dnl irix5.x)
dnl
2005-08-29 23:21:58 +08:00
AC_MSG_CHECKING([if accurately converting from integers to long double])
2005-07-23 09:55:12 +08:00
2005-08-29 23:21:58 +08:00
if test ${ac_cv_sizeof_long_double} = 0; then
hdf5_cv_integer_to_ldouble_accurate=${hdf5_cv_integer_to_ldouble_accurate=no}
else
AC_CACHE_VAL([hdf5_cv_integer_to_ldouble_accurate], [hdf5_cv_integer_to_ldouble_accurate=yes])
2005-07-23 09:55:12 +08:00
fi
2005-08-29 23:21:58 +08:00
if test ${hdf5_cv_integer_to_ldouble_accurate} = "yes"; then
AC_DEFINE([INTEGER_TO_LDOUBLE_ACCURATE], [1],
2005-01-09 04:56:12 +08:00
[Define if your system can accurately convert from integers to long double values.])
2005-01-07 02:17:59 +08:00
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2009-08-21 02:40:50 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
dnl 'unsigned long' to 'float' values.
dnl (This flag should be set for all machines, except for Pathscale compiler
dnl on Sandia's Linux machine where the compiler interprets 'unsigned long'
dnl values as negative when the first bit of 'unsigned long' is on during
dnl the conversion to float.)
dnl
AC_MSG_CHECKING([if accurately converting unsigned long to float values])
2009-09-01 00:21:56 +08:00
AC_CACHE_VAL([hdf5_cv_ulong_to_float_accurate],
2009-08-21 02:40:50 +08:00
[AC_TRY_RUN([
int main(void)
{
int ret = 0;
unsigned long l1;
unsigned long l2;
unsigned long l3;
float f1;
float f2;
float f3;
if(sizeof(unsigned long)==8) {
l1 = 0xffffffffffffffffUL;
l2 = 0xffffffffffff0000UL;
l3 = 0xf000000000000000UL;
f1 = (float)l1;
f2 = (float)l2;
f3 = (float)l3;
if((f1 < 0) || (f2 < 0) || (f3 < 0))
ret = 1;
}
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_ulong_to_float_accurate=yes], [hdf5_cv_ulong_to_float_accurate=no],)])
2009-08-21 02:40:50 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_ulong_to_float_accurate} = "yes"; then
2009-08-21 02:40:50 +08:00
AC_DEFINE([ULONG_TO_FLOAT_ACCURATE], [1],
[Define if your system accurately converting unsigned long to float values.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2004-04-29 01:02:12 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
2004-05-05 02:01:09 +08:00
dnl 'unsigned (long) long' values to 'float' and 'double' values.
2004-04-29 01:02:12 +08:00
dnl (This flag should be set for all machines, except for the SGIs, where
2004-05-05 02:01:09 +08:00
dnl the cache value is set in the config/irix6.x config file) and Solaris
dnl 64-bit machines, where the short program below tests if round-up is
dnl correctly handled.
2004-04-29 01:02:12 +08:00
dnl
2005-08-29 23:21:58 +08:00
AC_MSG_CHECKING([if accurately converting unsigned long long to floating-point values])
2004-05-05 02:01:09 +08:00
if test ${host_os_novers} = "solaris2.x"; then
2005-08-29 23:21:58 +08:00
AC_CACHE_VAL([hdf5_cv_ulong_to_fp_bottom_bit_accurate],
2004-05-05 02:01:09 +08:00
[AC_TRY_RUN([
int main(void)
{
unsigned long l1;
unsigned long l2;
unsigned long l3;
unsigned long l4;
unsigned long long ld1;
unsigned long long ld2;
unsigned long long ld3;
unsigned long long ld4;
double d1, d2, d3, d4;
unsigned char s[8];
int ret = 0;
if(sizeof(unsigned long)==8) {
2004-05-14 05:16:44 +08:00
l1 = 0xf000000000000b00UL; /*Round-down case*/
l2 = 0xf000000000000401UL; /*Round-up case*/
l3 = 0xf000000000000400UL; /*Round-down case*/
l4 = 0xf000000000000c00UL; /*Round-up case*/
2004-05-05 02:01:09 +08:00
d1 = (double)l1;
d2 = (double)l2;
d3 = (double)l3;
d4 = (double)l4;
} else if(sizeof(unsigned long long)==8) {
2004-05-14 05:16:44 +08:00
ld1 = 0xf000000000000b00ULL; /*Round-down case*/
ld2 = 0xf000000000000401ULL; /*Round-up case*/
ld3 = 0xf000000000000400ULL; /*Round-down case*/
ld4 = 0xf000000000000c00ULL; /*Round-up case*/
2004-05-05 02:01:09 +08:00
d1 = (double)ld1;
d2 = (double)ld2;
d3 = (double)ld3;
d4 = (double)ld4;
} else {
ret = 1;
goto done;
}
memcpy(s, &d1, 8);
if(s[7]!=1)
ret = 1;
memcpy(s, &d2, 8);
if(s[7]!=1)
ret = 1;
memcpy(s, &d3, 8);
if(s[7]!=0)
ret = 1;
memcpy(s, &d4, 8);
if(s[7]!=2)
ret = 1;
done:
exit(ret);
}
2005-08-29 23:21:58 +08:00
], [hdf5_cv_ulong_to_fp_bottom_bit_accurate=yes], [hdf5_cv_ulong_to_fp_bottom_bit_accurate=no],)])
2004-05-05 02:01:09 +08:00
else
2005-08-29 23:21:58 +08:00
AC_CACHE_VAL([hdf5_cv_ulong_to_fp_bottom_bit_accurate], [hdf5_cv_ulong_to_fp_bottom_bit_accurate=yes])
2004-05-05 02:01:09 +08:00
fi
2004-04-29 01:02:12 +08:00
2005-08-29 23:21:58 +08:00
if test ${hdf5_cv_ulong_to_fp_bottom_bit_accurate} = "yes"; then
AC_DEFINE([ULONG_TO_FP_BOTTOM_BIT_ACCURATE], [1],
2005-02-25 06:54:05 +08:00
[Define if your system can accurately convert unsigned (long) long values to floating-point values.])
2004-04-29 01:02:12 +08:00
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2004-05-08 05:58:54 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
2005-02-25 06:54:05 +08:00
dnl 'float' or 'double' to 'unsigned long long' values.
2004-05-08 05:58:54 +08:00
dnl (This flag should be set for all machines, except for PGI compiler
dnl where round-up happens when the fraction of float-point value is greater
dnl than 0.5.
dnl
2004-05-12 04:26:21 +08:00
AC_MSG_CHECKING([if accurately roundup converting floating-point to unsigned long long values])
2004-05-08 05:58:54 +08:00
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_fp_to_ullong_accurate],
2004-05-08 05:58:54 +08:00
[AC_TRY_RUN([
int main(void)
{
2004-05-14 05:16:44 +08:00
float f = 111.60f;
double d = 222.55L;
2004-05-08 05:58:54 +08:00
unsigned long long l1 = (unsigned long long)f;
unsigned long long l2 = (unsigned long long)d;
int ret = 0;
2004-05-12 04:26:21 +08:00
if(l1 == 112)
2004-05-08 05:58:54 +08:00
ret = 1;
2004-05-12 04:26:21 +08:00
if(l2 == 223)
2004-05-08 05:58:54 +08:00
ret = 1;
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_fp_to_ullong_accurate=yes], [hdf5_cv_fp_to_ullong_accurate=no],)])
2004-05-08 05:58:54 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_fp_to_ullong_accurate} = "yes"; then
2005-08-29 23:21:58 +08:00
AC_DEFINE([FP_TO_ULLONG_ACCURATE], [1],
[Define if your system roundup accurately converting floating-point to unsigned long long values.])
2004-05-08 05:58:54 +08:00
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2004-05-12 04:26:21 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
2005-02-25 06:54:05 +08:00
dnl 'float', 'double' or 'long double' to 'unsigned long long' values.
2004-05-12 04:26:21 +08:00
dnl (This flag should be set for all machines, except for HP-UX machines
dnl where the maximal number for unsigned long long is 0x7fffffffffffffff
dnl during conversion.
dnl
AC_MSG_CHECKING([if right maximum converting floating-point to unsigned long long values])
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_fp_to_ullong_right_maximum],
2004-05-12 04:26:21 +08:00
[AC_TRY_RUN([
int main(void)
{
float f = 9701917572145405952.00f;
2004-05-14 05:16:44 +08:00
double d1 = 9701917572145405952.00L;
2005-01-07 02:17:59 +08:00
long double d2 = 9701917572145405952.00L;
double d3 = 2e40L;
2004-05-12 04:26:21 +08:00
unsigned long long l1 = (unsigned long long)f;
2004-05-14 05:16:44 +08:00
unsigned long long l2 = (unsigned long long)d1;
2005-01-07 02:17:59 +08:00
unsigned long long l3 = (unsigned long long)d2;
unsigned long long l4;
2005-01-12 02:17:01 +08:00
unsigned long long l5 = 0x7fffffffffffffffULL;
2004-05-12 04:26:21 +08:00
int ret = 0;
2005-01-07 02:17:59 +08:00
if(l1 <= l5 || l2 <= l5 || l3 <= l5)
2004-05-14 05:16:44 +08:00
ret = 1;
2005-01-07 02:17:59 +08:00
l4 = (unsigned long long)d3;
if(l4 <= l5)
2004-05-12 04:26:21 +08:00
ret = 1;
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_fp_to_ullong_right_maximum=yes], [hdf5_cv_fp_to_ullong_right_maximum=no],)])
2004-05-12 04:26:21 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_fp_to_ullong_right_maximum} = "yes"; then
2004-05-12 04:26:21 +08:00
AC_DEFINE([FP_TO_ULLONG_RIGHT_MAXIMUM], [1],
[Define if your system has right maximum convert floating-point to unsigned long long values.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-01-15 04:15:34 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
dnl 'long double' to 'unsigned int' values. (This flag should be set for
dnl all machines, except for some Intel compilers on some Linux.)
dnl
2005-08-29 23:21:58 +08:00
AC_MSG_CHECKING([if correctly converting long double to unsigned int values])
2005-07-23 09:55:12 +08:00
2005-08-29 23:21:58 +08:00
if test ${ac_cv_sizeof_long_double} = 0; then
2009-10-16 05:14:27 +08:00
hdf5_cv_ldouble_to_uint_accurate=${hdf5_cv_ldouble_to_uint_accurate=no}
2005-08-29 23:21:58 +08:00
else
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_ldouble_to_uint_accurate],
2005-08-29 23:21:58 +08:00
[AC_TRY_RUN([
int main(void)
{
long double ld = 2733248032.9183987530L;
unsigned int i;
int ret = 0;
2005-01-15 04:15:34 +08:00
2005-08-29 23:21:58 +08:00
i = (unsigned int)ld;
if(i!=2733248032 && i!=2733248031 && i!=2733248033)
ret = 1;
2005-01-15 04:15:34 +08:00
2005-08-29 23:21:58 +08:00
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_ldouble_to_uint_accurate=yes], [hdf5_cv_ldouble_to_uint_accurate=no],)])
2005-08-29 23:21:58 +08:00
fi
2005-01-15 04:15:34 +08:00
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_ldouble_to_uint_accurate} = "yes"; then
2005-08-29 23:21:58 +08:00
AC_DEFINE([LDOUBLE_TO_UINT_ACCURATE], [1],
2005-01-15 04:15:34 +08:00
[Define if your system can convert long double to unsigned int values correctly.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2004-05-12 04:26:21 +08:00
2004-04-29 02:42:25 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can _compile_
dnl 'unsigned long long' to 'float' and 'double' typecasts.
dnl (This flag should be set for all machines, except for under Windows when
dnl compiled with Visual Studio 6, where the macro value is set in the
dnl src/H5pubconf.h file)
dnl
AC_MSG_CHECKING([if compiling unsigned long long to floating-point typecasts work])
AC_CACHE_VAL([hdf5_cv_ullong_to_fp_cast_works], [hdf5_cv_ullong_to_fp_cast_works=yes])
if test ${hdf5_cv_ullong_to_fp_cast_works} = "yes"; then
AC_DEFINE([ULLONG_TO_FP_CAST_WORKS], [1],
[Define if your system can compile unsigned long long to floating-point casts.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-04-30 05:39:05 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can _compile_
dnl 'long long' to 'float' and 'double' typecasts.
dnl (This flag should be set for all machines, except for under Windows when
dnl compiled with Visual Studio 6, where the macro value is set in the
dnl src/H5pubconf.h file)
dnl
AC_MSG_CHECKING([if compiling long long to floating-point typecasts work])
AC_CACHE_VAL([hdf5_cv_llong_to_fp_cast_works], [hdf5_cv_llong_to_fp_cast_works=yes])
if test ${hdf5_cv_llong_to_fp_cast_works} = "yes"; then
AC_DEFINE([LLONG_TO_FP_CAST_WORKS], [1],
[Define if your system can compile long long to floating-point casts.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-01-12 02:17:01 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can convert from
dnl 'unsigned long long' to 'long double' without precision loss.
dnl (This flag should be set for all machines, except for FreeBSD(sleipnir)
dnl where the last 2 bytes of mantissa are lost when compiler tries to do
2005-05-25 23:59:30 +08:00
dnl the conversion, and Cygwin where compiler doesn't do rounding correctly.)
2005-01-12 02:17:01 +08:00
dnl
2005-08-29 23:21:58 +08:00
AC_MSG_CHECKING([if converting unsigned long long to long double with precision])
2005-07-23 09:55:12 +08:00
2005-08-29 23:21:58 +08:00
if test ${ac_cv_sizeof_long_double} = 0; then
hdf5_cv_ullong_to_ldouble_precision=${hdf5_cv_ullong_to_ldouble_precision=no}
else
AC_CACHE_VAL([hdf5_cv_ullong_to_ldouble_precision],
[AC_TRY_RUN([
2010-04-28 05:35:53 +08:00
#include <stdlib.h>
#include <string.h>
2005-08-29 23:21:58 +08:00
int main(void)
{
/* General variables */
int endian;
int tst_value = 1;
int ret = 0;
/* For FreeBSD */
unsigned long long l = 0xa601e80bda85fcefULL;
long double ld;
unsigned char *c1, *c2;
size_t size;
/* For Cygwin */
unsigned long long l_cyg = 0xfffffffffffffff0ULL;
long double ld_cyg;
unsigned char *c2_cyg;
size_t size_cyg;
/* Determine this system's endianess */
c1 = (unsigned char*)calloc(1, sizeof(int));
memcpy((void*)c1, &tst_value, sizeof(int));
if(c1[0]==1)
endian = 0; /* little endian */
else
endian = 1; /* big endian */
/* For FreeBSD */
size = sizeof(long double);
memset(&ld, 0, size);
ld = (long double)l;
c2 = (unsigned char*)calloc(1, size);
memcpy((void*)c2, &ld, size);
/* Test if the last 2 bytes of mantissa are lost. Mainly for FreeBSD on Intel
* architecture(sleipnir) where it happens. */
/*if(endian==0 && c2[0]==0 && c2[1]==0)*/ /*little endian*/
if(endian==0 && c2[0]==0) { /*little endian*/
ret = 1;
goto done;
}
/* For Cygwin */
size_cyg = sizeof(long double);
memset(&ld_cyg, 0, size);
ld_cyg = (long double)l_cyg;
c2_cyg = (unsigned char*)calloc(1, size_cyg);
memcpy((void*)c2_cyg, &ld_cyg, size_cyg);
/* Test if the last 4 bytes(roughly) of mantissa are rounded up. Mainly for Cygwin
* where the values like 0xffffffffffffffff, 0xfffffffffffffffe, ...,
* 0xfffffffffffff000 ... are rounded up as 0x0000403f8000000000000000
* instead of 0x0000403effffffffffffffff, 0x0000403efffffffffffffffe, ...,
* 0x0000403efffffffffffff000 ...
*/
if(endian==0 && c2_cyg[0]==0 && c2_cyg[1]==0 && c2_cyg[2]==0 && c2_cyg[3]==0)
ret = 1;
done:
if(c1)
free(c1);
if(c2)
free(c2);
if(c2_cyg)
free(c2_cyg);
exit(ret);
2005-05-25 23:59:30 +08:00
}
2005-08-29 23:21:58 +08:00
], [hdf5_cv_ullong_to_ldouble_precision=yes], [hdf5_cv_ullong_to_ldouble_precision=no],)])
fi
2005-05-25 23:59:30 +08:00
2005-08-29 23:21:58 +08:00
if test ${hdf5_cv_ullong_to_ldouble_precision} = "yes"; then
AC_DEFINE([ULLONG_TO_LDOUBLE_PRECISION], [1],
2005-01-12 02:17:01 +08:00
[Define if your system can convert unsigned long long to long double with correct precision.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2005-07-20 02:16:33 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can handle overflow converting
dnl all floating-point to all integer types.
dnl (This flag should be set for all machines, except for Cray X1 where
dnl floating exception is generated when the floating-point value is greater
dnl than the maximal integer value).
dnl
AC_MSG_CHECKING([if overflows normally converting floating-point to integer values])
AC_CACHE_VAL([hdf5_cv_fp_to_integer_overflow_works],
[AC_TRY_RUN([
int main(void)
{
float f = 2147483648.0f;
int i;
i = (int)f;
done:
exit(0);
}
], [hdf5_cv_fp_to_integer_overflow_works=yes], [hdf5_cv_fp_to_integer_overflow_works=no],)])
if test ${hdf5_cv_fp_to_integer_overflow_works} = "yes"; then
AC_DEFINE([FP_TO_INTEGER_OVERFLOW_WORKS], [1],
[Define if your system can handle overflow converting floating-point to integer values.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2010-10-29 03:08:31 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine is using a special algorithm to convert
dnl 'long double' to '(unsigned) long' values. (This flag should only be set for
dnl the IBM Power6 Linux. When the bit sequence of long double is
dnl 0x4351ccf385ebc8a0bfcc2a3c3d855620, the converted value of (unsigned)long
dnl is 0x004733ce17af227f, not the same as the library's conversion to 0x004733ce17af2282.
dnl The machine's conversion gets the correct value. We define the macro and disable
dnl this kind of test until we figure out what algorithm they use.
dnl
AC_MSG_CHECKING([if using special algorithm to convert long double to (unsigned) long values])
if test ${ac_cv_sizeof_long_double} = 0; then
hdf5_cv_ldouble_to_long_special=${hdf5_cv_ldouble_to_long_special=no}
else
AC_CACHE_VAL([hdf5_cv_ldouble_to_long_special],
[AC_TRY_RUN([
int main(void)
{
long double ld = 20041683600089727.779961L;
long ll;
unsigned long ull;
unsigned char s[16];
unsigned char s2[8];
int ret = 1;
if(sizeof(long double) == 16 && sizeof(long) == 8) {
/*make sure the long double type has 16 bytes in size and
* 11 bits of exponent. If it is,
*the bit sequence should be like below. It's not
*a decent way to check but this info isn't available. */
memcpy(s, &ld, 16);
if(s[0]==0x43 && s[1]==0x51 && s[2]==0xcc && s[3]==0xf3 &&
s[4]==0x85 && s[5]==0xeb && s[6]==0xc8 && s[7]==0xa0 &&
s[8]==0xbf && s[9]==0xcc && s[10]==0x2a && s[11]==0x3c) {
/* Assign the hexadecimal value of long double type. */
s[0]=0x43; s[1]=0x51; s[2]=0xcc; s[3]=0xf3;
s[4]=0x85; s[5]=0xeb; s[6]=0xc8; s[7]=0xa0;
s[8]=0xbf; s[9]=0xcc; s[10]=0x2a; s[11]=0x3c;
s[12]=0x3d; s[13]=0x85; s[14]=0x56; s[15]=0x20;
memcpy(&ld, s, 16);
ll = (long)ld;
memcpy(s2, &ll, 8);
/* The library's algorithm converts it to 0x 00 47 33 ce 17 af 22 82
* and gets wrong value 20041683600089730 on the IBM Power6 Linux.
* But the IBM Power6 Linux converts it to 0x00 47 33 ce 17 af 22 7f
* and gets the correct value 20041683600089727. It uses some special
* algorithm. We're going to define the macro and skip the test until
* we can figure out how they do it. */
if(s2[0]==0x00 && s2[1]==0x47 && s2[2]==0x33 && s2[3]==0xce &&
s2[4]==0x17 && s2[5]==0xaf && s2[6]==0x22 && s2[7]==0x7f)
ret = 0;
ull = (unsigned long)ld;
memcpy(s2, &ull, 8);
/* The unsigned long is the same as signed long. */
if(s2[0]==0x00 && s2[1]==0x47 && s2[2]==0x33 && s2[3]==0xce &&
s2[4]==0x17 && s2[5]==0xaf && s2[6]==0x22 && s2[7]==0x7f)
ret = 0;
}
}
done:
exit(ret);
}
], [hdf5_cv_ldouble_to_long_special=yes], [hdf5_cv_ldouble_to_long_special=no],)])
fi
if test ${hdf5_cv_ldouble_to_long_special} = "yes"; then
AC_DEFINE([LDOUBLE_TO_LONG_SPECIAL], [1],
[Define if your system converts long double to (unsigned) long values with special algorithm.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine is using a special algorithm
dnl to convert some values of '(unsigned) long' to 'long double' values.
dnl (This flag should be off for all machines, except for IBM Power6 Linux,
dnl when the bit sequences are 003fff..., 007fff..., 00ffff..., 01ffff...,
dnl ..., 7fffff..., the compiler uses a unknown algorithm. We define a
dnl macro and skip the test for now until we know about the algorithm.
dnl
AC_MSG_CHECKING([if using special algorithm to convert (unsigned) long to long double values])
if test ${ac_cv_sizeof_long_double} = 0; then
hdf5_cv_long_to_ldouble_special=${hdf5_cv_long_to_ldouble_special=no}
else
AC_CACHE_VAL([hdf5_cv_long_to_ldouble_special],
[AC_TRY_RUN([
int main(void)
{
long double ld;
long ll;
unsigned long ull;
unsigned char s[16];
int flag=0, ret=1;
/*Determine if long double has 16 byte in size, 11 bit exponent, and
*the bias is 0x3ff */
if(sizeof(long double) == 16) {
ld = 1.0L;
memcpy(s, &ld, 16);
if(s[0]==0x3f && s[1]==0xf0 && s[2]==0x00 && s[3]==0x00 &&
s[4]==0x00 && s[5]==0x00 && s[6]==0x00 && s[7]==0x00)
flag = 1;
}
if(flag==1 && sizeof(long)==8) {
ll = 0x003fffffffffffffL;
ld = (long double)ll;
memcpy(s, &ld, 16);
/* The library converts the value to 0x434fffffffffffff8000000000000000.
* In decimal it is 18014398509481982.000000, one value short of the original.
* The IBM Power6 Linux converts it to 0x4350000000000000bff0000000000000.
* The value is correct in decimal. It uses some special
* algorithm. We're going to define the macro and skip the test until
* we can figure out how they do it. */
if(s[0]==0x43 && s[1]==0x50 && s[2]==0x00 && s[3]==0x00 &&
s[4]==0x00 && s[5]==0x00 && s[6]==0x00 && s[7]==0x00 &&
s[8]==0xbf && s[9]==0xf0 && s[10]==0x00 && s[11]==0x00 &&
s[12]==0x00 && s[13]==0x00 && s[14]==0x00 && s[15]==0x00)
ret = 0;
}
if(flag==1 && sizeof(unsigned long)==8) {
ull = 0xffffffffffffffffUL;
ld = (long double)ull;
memcpy(s, &ld, 16);
/* Use a different value from signed long to test. The problem is the same
* for both long and unsigned long. The value is 18446744073709551615.
* The library converts the value to 0x43effffffffffffffe000000000000000.
* In decimal it's 18446744073709548544.000000, very different from the original.
* The IBM Power6 Linux converts it to 0x43f0000000000000bff0000000000000.
* The value is correct in decimal. It uses some special
* algorithm. We're going to define the macro and skip the test until
* we can figure out how they do it. */
if(s[0]==0x43 && s[1]==0xf0 && s[2]==0x00 && s[3]==0x00 &&
s[4]==0x00 && s[5]==0x00 && s[6]==0x00 && s[7]==0x00 &&
s[8]==0xbf && s[9]==0xf0 && s[10]==0x00 && s[11]==0x00 &&
s[12]==0x00 && s[13]==0x00 && s[14]==0x00 && s[15]==0x00)
ret = 0;
}
done:
exit(ret);
}
], [hdf5_cv_long_to_ldouble_special=yes], [hdf5_cv_long_to_ldouble_special=no],)])
fi
if test ${hdf5_cv_long_to_ldouble_special} = "yes"; then
AC_DEFINE([LONG_TO_LDOUBLE_SPECIAL], [1],
[Define if your system can convert (unsigned) long to long double values with special algorithm.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2006-06-20 22:45:37 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
dnl 'long double' to '(unsigned) long long' values. (This flag should be set for
dnl all machines, except for Mac OS 10.4 and SGI IRIX64 6.5. When the bit sequence
dnl of long double is 0x4351ccf385ebc8a0bfcc2a3c..., the values of (unsigned)long long
dnl start to go wrong on these two machines. Adjusting it higher to
dnl 0x4351ccf385ebc8a0dfcc... or 0x4351ccf385ebc8a0ffcc... will make the converted
dnl values wildly wrong. This test detects this wrong behavior and disable the test.
dnl
AC_MSG_CHECKING([if correctly converting long double to (unsigned) long long values])
if test ${ac_cv_sizeof_long_double} = 0; then
2009-10-16 05:14:27 +08:00
hdf5_cv_ldouble_to_llong_accurate=${hdf5_cv_ldouble_to_llong_accurate=no}
2006-06-20 22:45:37 +08:00
else
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_ldouble_to_llong_accurate],
2006-06-20 22:45:37 +08:00
[AC_TRY_RUN([
int main(void)
{
long double ld = 20041683600089727.779961L;
long long ll;
unsigned long long ull;
unsigned char s[16];
int ret = 0;
if(sizeof(long double) == 16) {
/*make sure the long double type is the same as the failing type
*which has 16 bytes in size and 11 bits of exponent. If it is,
*the bit sequence should be like below. It's not
*a decent way to check but this info isn't available. */
memcpy(s, &ld, 16);
if(s[0]==0x43 && s[1]==0x51 && s[2]==0xcc && s[3]==0xf3 &&
s[4]==0x85 && s[5]==0xeb && s[6]==0xc8 && s[7]==0xa0 &&
s[8]==0xbf && s[9]==0xcc && s[10]==0x2a && s[11]==0x3c) {
/*slightly adjust the bit sequence (s[8]=0xdf). The converted
*values will go wild on Mac OS 10.4 and IRIX64 6.5.*/
s[0]=0x43; s[1]=0x51; s[2]=0xcc; s[3]=0xf3;
s[4]=0x85; s[5]=0xeb; s[6]=0xc8; s[7]=0xa0;
s[8]=0xdf; s[9]=0xcc; s[10]=0x2a; s[11]=0x3c;
s[12]=0x3d; s[13]=0x85; s[14]=0x56; s[15]=0x20;
memcpy(&ld, s, 16);
ll = (long long)ld;
ull = (unsigned long long)ld;
if(ll != 20041683600089728 || ull != 20041683600089728)
ret = 1;
}
}
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_ldouble_to_llong_accurate=yes], [hdf5_cv_ldouble_to_llong_accurate=no],)])
2006-06-20 22:45:37 +08:00
fi
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_ldouble_to_llong_accurate} = "yes"; then
2006-06-20 22:45:37 +08:00
AC_DEFINE([LDOUBLE_TO_LLONG_ACCURATE], [1],
[Define if your system can convert long double to (unsigned) long long values correctly.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2010-10-29 03:08:31 +08:00
2006-06-24 06:11:12 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine can accurately convert
dnl '(unsigned) long long' to 'long double' values. (This flag should be set for
dnl all machines, except for Mac OS 10.4, when the bit sequences are 003fff...,
dnl 007fff..., 00ffff..., 01ffff..., ..., 7fffff..., the converted values are twice
dnl as big as they should be.
dnl
AC_MSG_CHECKING([if correctly converting (unsigned) long long to long double values])
if test ${ac_cv_sizeof_long_double} = 0; then
2009-10-16 05:14:27 +08:00
hdf5_cv_llong_to_ldouble_correct=${hdf5_cv_llong_to_ldouble_correct=no}
2006-06-24 06:11:12 +08:00
else
[svn-r17345] Purpose:
Updating autotools
Description:
Installed new autotools and used them to reconfigure HDF5.
- Automake upgraded to 1.11
- Autoconf upgraded to 2.64
- bin/reconfigure script edited to use new versions (on jam), and
run to generate new configure script and Makefile.in's.
- configure.in script edited to add "_cv_" to all AC_CACHE_VAL strings
(in order to comply with new autoconf standard).
Tested:
Tested on machines jam, smirom, liberty, linew.
Tested w/ features c++, fortran, parallel.
Tested w/ compilers gcc, pgcc, icc.
Further testing via Daily Tests should catch any other outliers. Upon
passing DT's, I'll propogate the new tools into 1.8, hdf4, et cetera.
2009-08-13 02:55:25 +08:00
AC_CACHE_VAL([hdf5_cv_llong_to_ldouble_correct],
2006-06-24 06:11:12 +08:00
[AC_TRY_RUN([
int main(void)
{
long double ld;
long long ll;
unsigned long long ull;
unsigned char s[16];
int flag=0, ret=0;
/*Determine if long double has 16 byte in size, 11 bit exponent, and
*the bias is 0x3ff */
if(sizeof(long double) == 16) {
ld = 1.0L;
memcpy(s, &ld, 16);
if(s[0]==0x3f && s[1]==0xf0 && s[2]==0x00 && s[3]==0x00 &&
s[4]==0x00 && s[5]==0x00 && s[6]==0x00 && s[7]==0x00)
flag = 1;
}
if(flag==1 && sizeof(long long)==8) {
ll = 0x01ffffffffffffffLL;
ld = (long double)ll;
memcpy(s, &ld, 16);
/*Check if the bit sequence is as supposed to be*/
if(s[0]!=0x43 || s[1]!=0x7f || s[2]!=0xff || s[3]!=0xff ||
s[4]!=0xff || s[5]!=0xff || s[6]!=0xff || s[7]!=0xff ||
s[8]!=0xf0 || s[9]!=0x00 || s[10]!=0x00 || s[11]!=0x00)
ret = 1;
}
if(flag==1 && sizeof(unsigned long long)==8) {
ull = 0x01ffffffffffffffULL;
ld = (long double)ull;
memcpy(s, &ld, 16);
if(s[0]!=0x43 || s[1]!=0x7f || s[2]!=0xff || s[3]!=0xff ||
s[4]!=0xff || s[5]!=0xff || s[6]!=0xff || s[7]!=0xff ||
s[8]!=0xf0 || s[9]!=0x00 || s[10]!=0x00 || s[11]!=0x00)
ret = 1;
}
done:
exit(ret);
}
2009-10-16 05:14:27 +08:00
], [hdf5_cv_llong_to_ldouble_correct=yes], [hdf5_cv_llong_to_ldouble_correct=no],)])
2006-06-24 06:11:12 +08:00
fi
2009-10-16 05:14:27 +08:00
if test ${hdf5_cv_llong_to_ldouble_correct} = "yes"; then
2006-06-24 06:11:12 +08:00
AC_DEFINE([LLONG_TO_LDOUBLE_CORRECT], [1],
[Define if your system can convert (unsigned) long long to long double values correctly.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
2006-09-16 04:20:39 +08:00
dnl ----------------------------------------------------------------------
dnl Set the flag to indicate that the machine generates bad code
dnl for the H5V_log2_gen() routine in src/H5Vprivate.h
dnl (This flag should be set to no for all machines, except for SGI IRIX64,
dnl where the cache value is set to yes in it's config file)
dnl
AC_MSG_CHECKING([if bad code for log2 routine is generated])
AC_CACHE_VAL([hdf5_cv_bad_log2_code_generated], [hdf5_cv_bad_log2_code_generated=no])
if test ${hdf5_cv_bad_log2_code_generated} = "yes"; then
AC_DEFINE([BAD_LOG2_CODE_GENERATED], [1],
[Define if your system generates wrong code for log2 routine.])
AC_MSG_RESULT([yes])
else
AC_MSG_RESULT([no])
fi
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
dnl ----------------------------------------------------------------------
dnl Set some variables for general configuration information to be saved
dnl and installed with the libraries.
dnl
2001-07-06 04:56:01 +08:00
dnl HDF5 version from the first line of the README.txt file.
2002-03-21 02:14:36 +08:00
H5_VERSION="`cut -d' ' -f3 $srcdir/README.txt | head -1`"
2003-07-29 05:38:04 +08:00
AC_SUBST([H5_VERSION])
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
2000-09-15 06:19:24 +08:00
dnl Configuration date
2003-07-29 05:38:04 +08:00
AC_SUBST([CONFIG_DATE]) CONFIG_DATE="`date`"
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
2000-09-15 06:19:24 +08:00
dnl User doing the configuration
2003-07-29 05:38:04 +08:00
AC_SUBST([CONFIG_USER]) CONFIG_USER="`whoami`@`hostname`"
2000-09-15 06:19:24 +08:00
if test -n "$ORGANIZATION"; then
2001-09-28 07:05:16 +08:00
CONFIG_USER="$CONFIG_USER at $ORGANIZATION"
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
fi
2000-09-15 06:19:24 +08:00
dnl Configuration mode (production, development, profile, etc) saved above.
2003-07-29 05:38:04 +08:00
AC_SUBST([CONFIG_MODE])
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
2000-09-15 06:19:24 +08:00
dnl Byte sex from the AC_C_BIGENDIAN macro.
2003-07-29 05:38:04 +08:00
AC_SUBST([BYTESEX])
2000-09-15 06:19:24 +08:00
if test "X$ac_cv_c_bigendian" = "Xyes"; then
2001-09-28 07:05:16 +08:00
BYTESEX="big-endian"
2000-04-07 23:25:41 +08:00
else
2001-09-28 07:05:16 +08:00
BYTESEX="little-endian"
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
fi
2009-02-27 04:21:50 +08:00
if test "X$ac_cv_c_bigendian" = "Xyes"; then
WORDS_BIGENDIAN="yes"
else
WORDS_BIGENDIAN="no"
fi
AC_SUBST([WORDS_BIGENDIAN])
2000-09-15 06:19:24 +08:00
dnl Parallel support? (set above except empty if none)
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
PARALLEL=${PARALLEL:-no}
2000-11-16 22:50:38 +08:00
dnl Compiler with version information. This consists of the full path
dnl name of the compiler and the reported version number.
2003-07-29 05:38:04 +08:00
AC_SUBST([CC_VERSION])
2005-03-04 03:39:10 +08:00
dnl Strip anything that looks like a flag off of $CC
CC_NOFLAGS=`echo $CC | sed 's/ -.*//'`
if `echo $CC_NOFLAGS | grep ^/ >/dev/null 2>&1`; then
2005-03-02 02:47:23 +08:00
CC_VERSION="$CC"
2000-09-15 06:19:24 +08:00
else
2001-09-28 07:05:16 +08:00
CC_VERSION="$CC";
for x in `echo $PATH | sed -e 's/:/ /g'`; do
2008-05-14 00:43:42 +08:00
if test -x $x/$CC_NOFLAGS; then
2001-09-28 07:05:16 +08:00
CC_VERSION="$x/$CC"
break
fi
done
2000-11-16 22:50:38 +08:00
fi
2010-07-09 10:52:14 +08:00
if test -n "$cc_version_info"; then
CC_VERSION="$CC_VERSION ( $cc_version_info)"
fi
AC_SUBST([FC_VERSION])
dnl Strip anything that looks like a flag off of $CC
FC_NOFLAGS=`echo $FC | sed 's/ -.*//'`
if `echo $FC_NOFLAGS | grep ^/ >/dev/null 2>&1`; then
FC_VERSION="$FC"
else
FC_VERSION="$FC";
for x in `echo $PATH | sed -e 's/:/ /g'`; do
if test -x $x/$FC_NOFLAGS; then
FC_VERSION="$x/$FC"
break
fi
done
fi
if test -n "$fc_version_info"; then
FC_VERSION="$FC_VERSION ( $fc_version_info)"
fi
AC_SUBST([CXX_VERSION])
dnl Strip anything that looks like a flag off of $CC
CXX_NOFLAGS=`echo $CXX | sed 's/ -.*//'`
if `echo $CXX_NOFLAGS | grep ^/ >/dev/null 2>&1`; then
CXX_VERSION="$CXX"
else
CXX_VERSION="$FC";
for x in `echo $PATH | sed -e 's/:/ /g'`; do
if test -x $x/$CXX_NOFLAGS; then
CXX_VERSION="$x/$CXX"
break
fi
done
fi
if test -n "$cxx_version_info"; then
CXX_VERSION="$CXX_VERSION ( $cxx_version_info)"
2000-09-15 06:19:24 +08:00
fi
dnl ----------------------------------------------------------------------
2001-12-18 02:00:24 +08:00
dnl Where is the root of the source tree. Give an absolute address so
2000-09-15 06:19:24 +08:00
dnl we can find it no matter which directory of the distribution is our
2001-12-18 02:00:24 +08:00
dnl current directory. The built-in pwd fails on some systems, but the
2000-09-15 06:19:24 +08:00
dnl /bin/pwd version works OK.
dnl
if test -x /bin/pwd; then
2001-09-28 07:05:16 +08:00
pwd=/bin/pwd
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
else
2001-09-28 07:05:16 +08:00
pwd=pwd
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
fi
2003-07-29 05:38:04 +08:00
AC_SUBST([ROOT]) ROOT="`$pwd`"
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
2005-02-17 03:27:27 +08:00
dnl ----------------------------------------------------------------------
dnl Move any compiler-specific libraries into the main LIBS varaible.
dnl
LIBS="$DEFAULT_LIBS $LIBS"
2000-09-12 15:53:49 +08:00
dnl ----------------------------------------------------------------------
dnl Determine the runtime libraries we may need to include in the
dnl libtools command so that executables will find the correct dynamic
dnl libraries.
dnl
2003-07-29 05:38:04 +08:00
AC_SUBST([DYNAMIC_DIRS]) DYNAMIC_DIRS=""
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
if test -n "$AM_LDFLAGS $LDFLAGS"; then
for d in $AM_LDFLAGS $LDFLAGS ; do
2001-09-28 07:05:16 +08:00
case "$d" in
-L*)
2002-03-21 02:14:36 +08:00
d="`echo $d | sed -e 's/-L//g'`"
2001-09-28 07:05:16 +08:00
case "$d" in
.*)
dnl If the path isn't absolute, make it so by
dnl prepending the ROOT directory to it.
d=${ROOT}/$d
;;
esac
DYNAMIC_DIRS="-R${d} $DYNAMIC_DIRS"
;;
esac
done
2000-09-12 15:53:49 +08:00
fi
[svn-r1315] Changes since 19990607
----------------------
./configure.in
./configure [REGENERATED]
./src/libhdf5.settings.in [NEW]
./config/conclude.in
./INSTALL
./MANIFEST
A file called `libhdf5.settings' is installed along with the
libraries and it contains a list of various things that might
be of interest to someone using the library (such as what
compiler/options were used and what other libraries might need
to be linked to satisfy external references). We can add more
things as it becomes necessary. Here's a sample...
SUMMARY OF THE HDF5 CONFIGURATION
=================================
HDF5 Version: hdf5-1.1.106
Configured on: Tue Jun 8 11:38:52 EDT 1999
Configured by: matzke@llnl.gov at Spizella Software
Configure mode: development
Host system: i686-pc-linux-gnulibc1
Byte sex: little-endian
Libraries: static, shared
Parallel support: mpicc
Installation point: /usr/local
Compiler: mpicc (egcs-2.91.66)
Compiler switches: -Wall -g
Extra libraries: -lmfhdf -ldf -lz -ljpeg -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages: d,e,f,g,hg,i,mm,o,p,s,v,z
API Tracing: yes
File addresses: large
1999-06-09 00:14:10 +08:00
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
if test -n "$AM_CPPFLAGS"; then
2001-09-28 07:05:16 +08:00
TEMP_CPPFLAGS=""
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
for d in $AM_CPPFLAGS ; do
2001-09-28 07:05:16 +08:00
case "$d" in
-I.*)
dnl If the path isn't absolute, make it so by prepending
dnl the ROOT directory to it.
d="`echo $d | sed -e 's/-I//g'`"
d="-I${ROOT}/${d}"
;;
esac
TEMP_CPPFLAGS="$d $TEMP_CPPFLAGS"
done
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
AM_CPPFLAGS=$TEMP_CPPFLAGS
2000-09-15 06:19:24 +08:00
fi
2004-11-23 03:17:36 +08:00
dnl ----------------------------------------------------------------------
dnl Check if they would like the High Level library compiled
dnl
AC_SUBST(HL) HL=""
2004-12-09 04:31:25 +08:00
# name of fortran folder inside "hl", if FORTRAN compile is requested
AC_SUBST(HL_FOR) HL_FOR=""
2004-11-23 03:17:36 +08:00
AC_MSG_CHECKING([if high level library is enabled])
AC_ARG_ENABLE([hl],
2004-12-29 22:26:20 +08:00
[AC_HELP_STRING([--enable-hl],
[Enable the high level library [default=yes]])],
[HDF5_HL=$enableval],
[HDF5_HL=yes])
2004-11-23 03:17:36 +08:00
if test "X$HDF5_HL" = "Xyes"; then
echo "yes"
HL="hl"
2006-04-13 05:21:38 +08:00
AC_DEFINE([INCLUDE_HL], [1],
[Define if HDF5's high-level library headers should be included in hdf5.h])
2004-11-23 03:17:36 +08:00
else
echo "no"
fi
2005-10-15 06:52:13 +08:00
dnl ----------------------------------------------------------------------
dnl Some programs shouldn't be built by default (e.g., programs to generate
dnl data files used by tests, some optional tests).
dnl Check if they want such programs built anyway.
dnl
AC_MSG_CHECKING([additional programs should be built])
AC_ARG_ENABLE([build-all],
[AC_HELP_STRING([--enable-build-all],
[Build helper programs that only developers should need [default=no]])],
[BUILD_ALL=$enableval],
[BUILD_ALL=no])
if test "X$BUILD_ALL" = "Xyes"; then
echo "yes"
else
echo "no"
fi
AM_CONDITIONAL([BUILD_ALL_CONDITIONAL], [test "X$BUILD_ALL" = "Xyes"])
2007-07-13 06:01:35 +08:00
dnl ----------------------------------------------------------------------
dnl Enable deprecated public API symbols
dnl
AC_SUBST([DEPRECATED_SYMBOLS])
AC_MSG_CHECKING([if deprecated public symbols are available]);
AC_ARG_ENABLE([deprecated-symbols],
[AC_HELP_STRING([--enable-deprecated-symbols],
[Enable deprecated public API symbols [default=yes]])],
[DEPREC_SYMBOLS=$enableval],
[DEPREC_SYMBOLS=yes])
case "X-$DEPREC_SYMBOLS" in
X-yes)
AC_MSG_RESULT([yes])
DEPRECATED_SYMBOLS=yes
;;
X-no|*)
AC_MSG_RESULT([no])
DEPRECATED_SYMBOLS=no
AC_DEFINE([NO_DEPRECATED_SYMBOLS], [1],
[Define if deprecated public API symbols are disabled])
;;
esac
2007-07-13 10:42:17 +08:00
dnl --------------------------------------------------------------------------
dnl Which version of the public APIs should the 'base' versioned symbols use?
dnl
AC_SUBST([DEFAULT_API_VERSION])
AC_MSG_CHECKING([which version of public symbols to use by default])
AC_ARG_WITH([default-api-version],
[svn-r17582] Description:
Bring changes from file free space branch back to the trunk. *yay!*
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.5.8 (amazon) in debug mode
Mac OS X/32 10.5.8 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2009-10-02 10:08:59 +08:00
[AC_HELP_STRING([--with-default-api-version=(v16|v18|v110)],
2007-07-13 10:42:17 +08:00
[Specify default release version of public symbols
[svn-r17582] Description:
Bring changes from file free space branch back to the trunk. *yay!*
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.5.8 (amazon) in debug mode
Mac OS X/32 10.5.8 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2009-10-02 10:08:59 +08:00
[default=v110]])],,
withval=v110)
2007-07-13 10:42:17 +08:00
if test "X$withval" = "Xv16"; then
AC_MSG_RESULT([v16])
DEFAULT_API_VERSION=v16
2008-10-25 03:12:21 +08:00
AC_DEFINE([USE_16_API_DEFAULT], [1],
2007-07-13 10:42:17 +08:00
[Define using v1.6 public API symbols by default])
elif test "X$withval" = "Xv18"; then
AC_MSG_RESULT([v18])
DEFAULT_API_VERSION=v18
[svn-r17582] Description:
Bring changes from file free space branch back to the trunk. *yay!*
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/default API=1.8.x,
w/C++ & FORTRAN, w/threadsafe, in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers, w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.5.8 (amazon) in debug mode
Mac OS X/32 10.5.8 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
2009-10-02 10:08:59 +08:00
elif test "X$withval" = "Xv110"; then
AC_MSG_RESULT([v110])
DEFAULT_API_VERSION=v110
2007-07-13 10:42:17 +08:00
else
AC_MSG_ERROR([invalid version of public symbols given])
fi
dnl It's an error to try to disable deprecated public API symbols while
2010-11-18 02:07:51 +08:00
dnl choosing an older version of the public API as the default. However,
dnl if the user insists on doing this via the --enable-unsupported configure
dnl flag, we'll let them.
if test "X${ALLOW_UNSUPPORTED}" != "Xyes"; then
if test "X${DEFAULT_API_VERSION}" != "Xv110" -a "X${DEPRECATED_SYMBOLS}" = "Xno" ; then
AC_MSG_ERROR([Removing old public API symbols not allowed when using them as default public API symbols. Use --enable-unsupported to override this error.])
fi
2007-07-13 10:42:17 +08:00
fi
[svn-r14402] Description:
Add work-around to allow reading files that were produced with a buggy
earlier version of the library, which could create objects with the wrong
object header message count. There is now a configure flag
"--enable-strict-format-checks" which triggers a failure on reading a file
with this sort of corruption (when enabled) and allows the object to be read
(when disabled). The default value for the "strict-format-checks" flag is
yes when the "debug" flag is enabled and no when the "debug" flag is disabled.
Note that if strict format checks are disabled (allowing objects with
this particular kind of corruption to be read) and the file is opened with
write access, the library will re-write the object header for the corrupt
object with the correct # of object header messages.
This closes bugzilla bug #1010.
Tested on:
FreeBSD/32 6.2 (duty) in debug mode
FreeBSD/64 6.2 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (kagiso) w/PGI compilers, w/C++ & FORTRAN, w/threadsafe,
in debug mode
Linux/64-amd64 2.6 (smirom) w/default API=1.6.x, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Mac OS X/32 10.4.10 (amazon) in debug mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in production mode
2008-01-13 13:37:00 +08:00
dnl ----------------------------------------------------------------------
dnl Enable strict file format checks
dnl
AC_SUBST([STRICT_FORMAT_CHECKS])
AC_MSG_CHECKING([Whether to perform strict file format checks]);
AC_ARG_ENABLE([strict-format-checks],
[AC_HELP_STRING([--enable-strict-format-checks],
[Enable strict file format checks, default=yes if
debug flag is enabled, no otherwise])],
[STRICT_CHECKS=$enableval])
dnl Default to yes if debug is enabled
if test "X-$STRICT_CHECKS" = X- ; then
if test -z "$DEBUG_PKG" ; then
STRICT_CHECKS=no
else
STRICT_CHECKS=yes
fi
fi
case "X-$STRICT_CHECKS" in
X-yes)
AC_MSG_RESULT([yes])
STRICT_FORMAT_CHECKS=yes
AC_DEFINE([STRICT_FORMAT_CHECKS], [1],
[Define if strict file format checks are enabled])
;;
X-no|*)
AC_MSG_RESULT([no])
STRICT_FORMAT_CHECKS=no
;;
esac
2009-04-21 10:02:46 +08:00
dnl ----------------------------------------------------------------------
dnl Enable embedded library information
dnl
AC_MSG_CHECKING([Whether to have library information embedded in the executables])
AC_ARG_ENABLE([embedded-libinfo],
[AC_HELP_STRING([--enable-embedded-libinfo],
[Enable embedded library information [default=yes]])],
[enable_embedded_libinfo=$enableval],
[enable_embedded_libinfo=yes])
if test "${enable_embedded_libinfo}" = "yes"; then
AC_MSG_RESULT([yes])
AC_DEFINE([HAVE_EMBEDDED_LIBINFO], [1],
[Define if library information should be embedded in the executables])
else
AC_MSG_RESULT([no])
fi
2008-09-19 04:53:30 +08:00
dnl ----------------------------------------------------------------------
dnl Check if pointer alignments are enforced
dnl
AC_MSG_CHECKING([if alignment restrictions are strictly enforced])
AC_RUN_IFELSE([
AC_LANG_PROGRAM([
#include <stdlib.h>
#include <string.h>
typedef struct {
size_t len;
void *p;
} hvl_t;
], [
char *chp = "beefs";
char **chpp = malloc (2 * sizeof (char *));
char **chpp2;
hvl_t vl = { 12345, (void *) chp };
hvl_t *vlp;
hvl_t *vlp2;
memcpy ((void *) ((char *) chpp + 1), &chp, sizeof (char *));
chpp2 = (char **) ((char *) chpp + 1);
if (strcmp (*chpp2, chp)) {
free (chpp);
return 1;
}
free (chpp);
vlp = malloc (2 * sizeof (hvl_t));
memcpy ((void *) ((char *) vlp + 1), &vl, sizeof (hvl_t));
vlp2 = (hvl_t *) ((char *) vlp + 1);
if (vlp2->len != vl.len || vlp2->p != vl.p) {
free (vlp);
return 1;
}
free (vlp);
])
], [
AC_DEFINE([NO_ALIGNMENT_RESTRICTIONS], [1], [Define if we can violate pointer alignment restrictions])
AC_MSG_RESULT([no])
], [
AC_MSG_RESULT([yes])
], [
AC_MSG_RESULT([unknown, assuming yes])
])
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
2009-10-19 11:05:25 +08:00
dnl ----------------------------------------------------------------------
dnl Restore user's CFLAGS.
2009-10-14 05:55:21 +08:00
CFLAGS="$saved_user_CFLAGS"
2009-10-19 11:05:25 +08:00
FCFLAGS="$saved_user_FCFLAGS"
2009-10-14 05:55:21 +08:00
CXXFLAGS="$saved_user_CXXFLAGS"
2009-10-19 11:05:25 +08:00
CPPFLAGS="$saved_user_CPPFLAGS"
LDFLAGS="$saved_user_LDFLAGS"
[svn-r17616] Purpose:
CFLAGS overhaul
Description:
Modified the way configure handles CFLAGS.
(note: all changes also apply to FCFLAGS, CXXFLAGS, CPPFLAGS, and LDFLAGS).
1. The configure process will now always preserve a user's CFLAGS
environment variable setup. Any additional flags necessary for compilation
added at configure time will be passed into the Makefiles as AM_CFLAGS,
which is an automake construct to be used in addition to CFLAGS.
This will allow a user to have the final say, as CFLAGS will always appear
later in the compile line than AM_CFLAGS. Additionally, setting CFLAGS
during make will no longer completely erase all flags set by configure,
since they're maintained in AM_CFLAGS.
2. Additionally, where possible, flags previously being assigned directly
into CFLAGS (and thus propagating into h5cc) have now been redirected into
H5_CFLAGS, so they're used ONLY for compiling hdf5, and not embedded into
the h5cc wrapper script as well.
*Note that H5_CFLAGS ultimately is assigned into AM_CFLAGS for use in the
Makefiles. Complete description of changes and build process will be
included in a Configure Document that Elena and I are working on.
3. Removed unsupported config files. This includes:
config/dec-osf*
config/hpux11.00
config/irix5.x
config/powerpc-ibm-aix4.x
config/rs6000-ibm-aix5.x
config/unicos*
4. Modified configure summary to display additional values. Specifically,
appropriate AM_* variables are being shown, as well as H5_FCFLAGS and
H5_CXXFLAGS, which were for some reason not already present.
Tested:
- H5committest
- Tested on all THG / NCSA machines, using several combinations of the more
prominent configure options (c++, fortran, szip, threadsafe, parallel,
et cetera). (Thanks to Quincey for rysnc testing setup!)
- With regards to new automated testing, anything *necessary* for
compilation will be caught by the daily tests as it stands now. (i.e.,
if LDFLAGS is not properly set when szip is used, linking will fail).
Additionally, with regards to which flags get into h5cc, if any
*necessary* flags have been improperly removed, then daily tests should
fail during make installcheck. Additional machine-specific desired
behaviors and/or checks may have to be set up separately within the
daily tests, so this is something to work on.
2009-10-08 23:17:30 +08:00
1997-08-16 00:35:53 +08:00
dnl ----------------------------------------------------------------------
2005-02-01 11:17:02 +08:00
dnl Create automake conditionals to tell automake makefiles which directories
dnl need to be compiled
AM_CONDITIONAL([BUILD_CXX_CONDITIONAL], [test "X$HDF_CXX" = "Xyes"])
AM_CONDITIONAL([BUILD_PARALLEL_CONDITIONAL], [test -n "$TESTPARALLEL"])
AM_CONDITIONAL([BUILD_FORTRAN_CONDITIONAL], [test "X$HDF_FORTRAN" = "Xyes"])
AM_CONDITIONAL([BUILD_HDF5_HL_CONDITIONAL], [test "X$HDF5_HL" = "Xyes"])
2006-06-22 07:15:08 +08:00
2005-02-01 11:17:02 +08:00
dnl ----------------------------------------------------------------------
dnl Build the Makefiles.
1997-08-16 00:35:53 +08:00
dnl
1999-11-01 23:21:16 +08:00
dnl The directory search list
2003-07-29 05:38:04 +08:00
AC_SUBST([SEARCH]) SEARCH='$(srcdir) $(top_builddir)/src $(top_srcdir)/src'
1999-11-01 23:21:16 +08:00
cmd='echo $SEARCH |sed "s/ /'$SEARCH_SEP'/g"'
2000-10-12 01:25:21 +08:00
SEARCH="$SEARCH_RULE`eval $cmd`"
export SEARCH
1997-08-16 00:35:53 +08:00
dnl We don't need to say when we're entering directories if we're using
2003-08-09 02:28:03 +08:00
dnl GNU make because make does it for us.
1997-08-16 00:35:53 +08:00
if test "X$GMAKE" = "Xyes"; then
2003-07-29 05:38:04 +08:00
AC_SUBST([SETX]) SETX=":"
1997-08-16 00:35:53 +08:00
else
2003-07-29 05:38:04 +08:00
AC_SUBST([SETX]) SETX="set -x"
1997-08-16 00:35:53 +08:00
fi
1999-03-20 04:09:50 +08:00
dnl Some cleanup stuff
2009-10-16 02:11:56 +08:00
rm -f conftest conftest.o conftest.c dummy.o *.mod
1997-08-29 00:36:57 +08:00
1999-03-23 00:21:49 +08:00
dnl Build config.status, touch the stamp files, and build all the Makefiles.
dnl The order is such that the first `make' does not need to update any
dnl configuration information. See config/commence.in for the order in which
dnl things need to be done.
# First the stamp1 file for H5config.h.in
mkdir ./config >/dev/null 2>&1
touch ./config/stamp1
# Then the config.status file (but not makefiles)
saved_no_create=$no_create
1999-03-20 04:09:50 +08:00
no_create=yes
2001-02-23 06:02:01 +08:00
2001-12-18 02:00:24 +08:00
PARALLEL_MAKE=""
2003-07-29 05:38:04 +08:00
FORTRAN_PARALLEL_MAKE=""
2001-04-03 04:33:14 +08:00
if test -n "$TESTPARALLEL"; then
2002-07-15 23:17:05 +08:00
PARALLEL_MAKE="$TESTPARALLEL/Makefile $TESTPARALLEL/testph5.sh"
2003-07-29 05:38:04 +08:00
if test "X$HDF_FORTRAN" = "Xyes"; then
FORTRAN_PARALLEL_MAKE=fortran/$TESTPARALLEL/Makefile
fi
2001-04-03 04:33:14 +08:00
fi
2002-04-23 07:44:41 +08:00
AC_CONFIG_FILES([src/libhdf5.settings
Makefile
src/Makefile
test/Makefile
2009-09-28 23:44:35 +08:00
test/testcheck_version.sh
2007-08-22 05:08:27 +08:00
test/testerror.sh
2010-03-18 20:23:04 +08:00
test/H5srcdir_str.h
2009-09-21 13:28:44 +08:00
test/testlibinfo.sh
2010-11-15 12:04:28 +08:00
test/testlinks_env.sh
2005-02-01 11:17:02 +08:00
testpar/Makefile
testpar/testph5.sh
2002-04-23 07:44:41 +08:00
perform/Makefile
tools/Makefile
tools/h5dump/Makefile
2004-06-09 21:47:20 +08:00
tools/h5dump/testh5dump.sh
2006-06-14 05:21:14 +08:00
tools/h5dump/testh5dumpxml.sh
2009-02-27 04:21:50 +08:00
tools/h5ls/testh5ls.sh
2003-04-04 04:00:42 +08:00
tools/h5import/Makefile
2002-10-10 05:25:48 +08:00
tools/h5diff/Makefile
2004-12-29 22:26:20 +08:00
tools/h5jam/Makefile
tools/h5jam/testh5jam.sh
2003-09-20 03:32:25 +08:00
tools/h5repack/Makefile
2004-07-02 01:38:04 +08:00
tools/h5repack/h5repack.sh
2002-04-23 07:44:41 +08:00
tools/h5ls/Makefile
2006-07-20 05:47:33 +08:00
tools/h5copy/Makefile
2002-04-23 07:44:41 +08:00
tools/lib/Makefile
tools/misc/Makefile
tools/misc/h5cc
2005-06-03 05:00:23 +08:00
tools/misc/testh5repart.sh
2007-07-28 01:36:36 +08:00
tools/h5stat/testh5stat.sh
tools/h5stat/Makefile
2002-04-23 07:44:41 +08:00
examples/Makefile
2010-07-09 10:52:14 +08:00
examples/run-c-ex.sh
2007-04-11 13:17:13 +08:00
examples/testh5cc.sh
2005-02-01 11:17:02 +08:00
c++/Makefile
c++/src/Makefile
c++/src/h5c++
c++/test/Makefile
2010-03-18 20:23:04 +08:00
c++/test/H5srcdir_str.h
2005-02-01 11:17:02 +08:00
c++/examples/Makefile
2010-07-09 10:52:14 +08:00
c++/examples/run-c++-ex.sh
2007-04-11 13:17:13 +08:00
c++/examples/testh5c++.sh
2005-02-01 11:17:02 +08:00
fortran/Makefile
fortran/src/h5fc
fortran/src/Makefile
fortran/test/Makefile
fortran/testpar/Makefile
fortran/examples/Makefile
2010-07-09 10:52:14 +08:00
fortran/examples/run-fortran-ex.sh
2007-04-11 13:17:13 +08:00
fortran/examples/testh5fc.sh
2005-02-01 11:17:02 +08:00
hl/Makefile
hl/src/Makefile
hl/test/Makefile
2010-03-18 20:23:04 +08:00
hl/test/H5srcdir_str.h
2006-03-04 05:49:10 +08:00
hl/tools/Makefile
hl/tools/gif2h5/Makefile
2005-12-10 04:59:11 +08:00
hl/examples/Makefile
2010-07-09 10:52:14 +08:00
hl/examples/run-hlc-ex.sh
2005-03-29 00:32:20 +08:00
hl/c++/Makefile
hl/c++/src/Makefile
hl/c++/test/Makefile
2005-12-10 04:59:11 +08:00
hl/c++/examples/Makefile
2010-07-09 10:52:14 +08:00
hl/c++/examples/run-hlc++-ex.sh
2005-02-01 11:17:02 +08:00
hl/fortran/Makefile
hl/fortran/src/Makefile
2005-12-10 04:59:11 +08:00
hl/fortran/test/Makefile
2010-07-09 10:52:14 +08:00
hl/fortran/examples/Makefile
hl/fortran/examples/run-hlfortran-ex.sh])
2002-04-23 07:44:41 +08:00
AC_OUTPUT
2011-04-15 05:21:59 +08:00
LT_OUTPUT
1999-03-23 00:21:49 +08:00
no_create=$saved_no_create
# Then the stamp2 file for H5config.h
touch ./config/stamp2
# Finally the makefiles
test "$no_create" = yes || ${CONFIG_SHELL-/bin/sh} $CONFIG_STATUS || exit 1
2000-09-07 02:31:47 +08:00
2008-05-06 02:35:55 +08:00
dnl Are we compiling static libraries, shared libraries, or both? This
dnl is only used for the libhdf5.settings file. We can't just look at
dnl $enable_static and $enable_shared because if they're yes the ltconfig
dnl might have decided that one or the other is simply not possible.
dnl Therefore we have to ask the generated `libtool' shell script
dnl which 'features' it has enabled.
if (./libtool --features | grep '^enable shared libraries' > /dev/null); then
enable_shared=yes
else
enable_shared=no
fi
if (./libtool --features | grep '^enable static libraries' > /dev/null); then
enable_static=yes
else
enable_static=no
fi
if test "X$enable_static" = "Xyes" && test "X$enable_shared" = "Xyes"; then
STATIC_SHARED="static, shared"
elif test "X$enable_static" = "Xyes"; then
STATIC_SHARED="static"
elif test "X$enable_shared" = "Xyes"; then
STATIC_SHARED="shared"
else
STATIC_SHARED="none"
fi
2002-03-08 04:07:40 +08:00
chmod 755 tools/misc/h5cc
2003-07-29 05:38:04 +08:00
if test "X$HDF_FORTRAN" = "Xyes"; then
chmod 755 fortran/src/h5fc
fi
if test "X$HDF_CXX" = "Xyes"; then
chmod 755 c++/src/h5c++
fi
2000-09-07 02:31:47 +08:00
dnl We don't want inline defined for C++ compilers
2001-07-18 05:24:38 +08:00
dnl Don't worry about the C++ ifdef wrappers in the H5pubconf file, since
dnl 'H5_inline' isn't a C++ keyword.
2000-09-07 02:31:47 +08:00
cat >> src/H5config.h <<EOF
2000-09-08 04:57:34 +08:00
#if defined(__cplusplus) && defined(inline)
#undef inline
2000-09-07 02:31:47 +08:00
#endif
EOF
2008-05-15 13:12:00 +08:00
dnl show the configure settings
cat src/libhdf5.settings